Master Art of Translation | The Multilingual Comprehension Testing System — How Cognitive Interviews, Teach-Back, Reader Tasks and Retesting Validate Meaning

Architecture: EDKSG-TRANS-MASTER-WORLD-1220. This chapter addresses reader-understanding validation as a separate discipline from translation drafting, automated string tests, generic language-quality scoring and post-release error reporting.

How do you test whether a translation is actually understood? A reliable multilingual comprehension test does more than ask a fluent bilingual reviewer whether the words sound natural. It asks the intended reader to interpret a translated message, make the decision the message requires, explain a condition without being coached, and show what action they would take. Cognitive interviews, teach-back, task-based usability tests and evidence-led retesting help expose failures that ordinary proofreading cannot see.

People searching for translation comprehension testing, cognitive debriefing of translated questionnaires, linguistic validation, multilingual usability testing or plain-language testing usually have a practical problem: the translated text appears correct, yet users misunderstand deadlines, permissions, exceptions, amounts or next steps. The useful question is not simply whether a sentence has equivalent words. It is whether the new reader reconstructs the intended situation, constraints and choice.

This master guide explains how to validate translated instructions and multilingual content with real reader tasks, neutral probes, interpretation maps, controlled tests, error attribution and documented release decisions. It supplies a repeatable system, diagnostic cases across ordinary services and higher-consequence situations, and routes to related work without replacing specialist translation quality, accessibility or vocabulary resources.

Quick answer: a twelve-step reader-comprehension test

1. Define the real decision or behavior the reader must perform, not merely the topic of the text.
2. Identify the source claim, actor, condition, exception, time, quantity, consequence and uncertainty.
3. Separate what the organization means from how the organization happens to word it.
4. Recruit readers who resemble the intended audience in language proficiency and task familiarity.
5. Keep an unprompted first exposure to the translated artifact.
6. Ask an open paraphrase before any leading follow-up.
7. Observe a realistic task, such as selecting a safe option or locating the correct next step.
8. Probe boundary conditions and misleading interpretations with contrasting examples.
9. Record exact reader actions and explanations, not a facilitator’s impression of success.
10. Diagnose whether the problem comes from the source, translation, interface, context or prior knowledge.
11. Revise the smallest necessary unit and check consistency with the source’s intent.
12. Retest the corrected material using comparable tasks and independent evidence.

Search intent: why translation accuracy is not enough

A translation can be lexically accurate but behaviorally misleading. Suppose a notice says, “You may cancel before the renewal date; cancellation takes effect at the end of the current period.” A reader may correctly recognize every word and still conclude that cancellation immediately ends access. A spellchecker, grammar check and even a clean back translation may fail to expose that inference. The comprehension question is whether the reader can predict both the billing consequence and the period of remaining access.

Another reader may understand the sentence but fail because the button is labeled with a false friend, the date is shown in an unfamiliar format or the exception lives behind a collapsed panel. This guide treats those as different causes with different fixes. Replacing a translation when the interface hides a condition can make the sentence longer without making the decision safer. Likewise, changing the interface cannot compensate for a translated negation that reverses permission.

The hidden problem is that fluent phrasing, source fidelity, linguistic equivalence, usability, perceived confidence and actual comprehension are separate dimensions. A robust program measures them separately. The reader’s correct action is the central outcome, but it does not replace professional review, accessibility requirements or domain sign-off. In consequential contexts, a small qualitative study diagnoses risk; it does not mathematically prove that every future reader will act correctly.

Research basis and authority: what the sources actually support

The ISO 24495-1:2023 plain-language standard sets governing principles for developing plain-language documents across most written languages. Its public summary says that the standard focuses on written content and does not itself supply all digital accessibility guidance. In this chapter, its role is a writing and evaluation foundation, not a claim that one index certifies any translation as understood.

The CDC Clear Communication Index guidance, checked against the live source for this chapter, recommends testing material with the primary audience even after independent reviews. It distinguishes reviewer scoring from confirmation that actual readers understand and can use the material. Its published guidance is a communication reference, not a universal minimum sample size for every language, domain or product.

The ISPOR Task Force principles for translation and cultural adaptation identify cognitive debriefing among the steps used in a particular patient-reported-outcomes context. The original 2005 guidance discusses small groups of relevant respondents; it is not authority for imposing one universal participant count on public notices, software interfaces or all education materials. Borrow the underlying principle—test interpretation with the intended audience—while designing sampling for the actual risk and readership.

The IEC/IEEE 82079-1:2019 guidance on information for use expressly addresses preparation and empirical evaluation of instructions for use. It is useful when readers must follow a sequence or use a product, but the existence of a standard does not remove the need to verify the applicable regulatory and professional context. For digital interfaces, the W3C WCAG 2.2 Understanding resources help separate language, labels, instructions and accessibility failure modes.

Part I. The mechanism: from source sentence to reader decision

1. Build an interpretation contract before writing a test

An interpretation contract is a short statement of the meaning that must survive translation. It does not dictate target-language syntax. It records the actor, permitted or required action, triggering condition, temporal order, quantitative threshold, exception, evidence state and consequence. When two source sentences imply different actions, they must not be collapsed merely because the same familiar word appears in both.

The contract is useful because a facilitator otherwise invents a correct answer after watching the reader. Prepare it before exposing anyone to the text. Let a subject specialist verify the intended decision; let a translator verify linguistic realization; let the reader reveal whether those combine into comprehensible communication. A contract is not a second translation—it is an explicit set of invariants.

Write an example with two possible interpretations. For a cancellation notice, one possible interpretation is that access ends immediately, while another is that access remains through the paid period. Test the difference instead of asking whether the sentence is clear. For any message with legal or safety implications, verify the approved source meaning with its accountable owner before judging the target text.

2. Define what the reader can reasonably know

A reader does not approach a message as an empty dictionary. They bring habits from similar services, assumptions about authority, familiarity with symbols, prior failures and knowledge of social conventions. These expectations can help them infer missing details; they can also push them toward a confident but incorrect interpretation. The same translation may therefore produce different behavior across reader populations.

Document the entry point and the information visible at that moment. Did the reader arrive from an email, a shared link, a QR code, a printed form or a search result? Is there a related label, button, status icon or help text? A test that removes all surrounding context may overestimate difficulty. A test that provides a facilitator’s explanation may hide a real-world problem.

Recruit for the task rather than for abstract fluency alone. Language proficiency matters, but so do digital literacy, familiarity with the process, disabilities affecting use, professional expertise and whether the reader is interpreting a personal consequence. Avoid treating any person as a representative of an entire language group.

3. Separate comprehension, recall, trust and completion

A reader can accurately repeat a translated notice and still fail to carry out the task. Conversely, someone may complete a familiar task by guessing the standard button positions while misreading the conditions. Comprehension, memory, confidence, task success and agreement with the policy are separate observations. A good study records each rather than compressing them into a single satisfaction score.

Paraphrase tests conceptual interpretation. Task simulation tests action. Recall tests retention after distraction. Confidence ratings describe subjective certainty. A direct consent or acceptance response describes willingness, not necessarily understanding. None of these measures substitutes for the others, and their disagreements are diagnostically valuable.

Define the evidence before the session. For example, success may require identifying both the cancellation deadline and the continued-access condition. If the participant clicks the correct control but predicts an incorrect outcome, report partial success rather than a pass. Describe uncertainty honestly instead of reporting spurious precision.

4. Why back translation is insufficient as a sole gate

Back translation can expose some discrepancies, especially when a target version changes a technical concept or omits a clause. But reproducing source-like wording does not demonstrate that ordinary readers will understand it. Different target expressions can back-translate similarly while carrying different implications about permission, politeness, urgency or familiarity.

The same issue appears with expert consensus. Two translators may agree that a term is defensible while intended readers interpret it as a different stage of a process. Neither the expert judgment nor the reader response should be discarded: the tension is a finding. The source author may need to clarify the original, the translator may need a more explicit rendering, or the interface may need to show the condition at the point of choice.

Use back translation as one possible diagnostic instrument under an appropriate process, not as a badge of universal validation. In regulated instruments, follow the actual governing protocol. In everyday public information, focus the study on the reader’s decision and preserve records of why a revision was accepted.

5. Build an unbiased first-exposure test

Facilitators easily leak the answer through emphasis, facial expression, voice intonation or the way they paraphrase a question. Asking “You understand that this is refundable, right?” measures compliance with the facilitator more than independent comprehension. Even “What does refundable mean?” may prime a reader who had not previously noticed that concept.

Begin with an unprompted task such as “Tell me what happens next” or “Show me what you would do at this point.” Record what the reader says before supplying clarification. Follow with neutral probes that inspect condition, time, responsibility and consequence. Interviewers should use a stable script but have room to ask for a concrete example.

Mark answers given before and after prompting separately. A reader who corrects themselves after a leading explanation has not provided the same evidence as one who understood unaided. Keep the prompt transcript so a reviewer can distinguish a genuine correction from an interviewer-induced one.

6. Design tasks around decisions, not trivia

If the message governs a decision, a test asking for a dictionary definition may be irrelevant. Someone can define “authorization” yet select the wrong permission scope. Build a scenario with enough realistic stakes to reveal the interpretation: an account has an upcoming charge, a document contains two versions, or a form asks whether an applicant qualifies for an exception.

A useful task has an observable outcome, a plausible wrong branch and a reason why the wrong branch matters. Avoid unnecessary personal data. Use invented accounts, sample records and clearly fictional fixtures. Never ask participants to make real financial, medical or safety decisions as part of a content test.

Task success includes the path, not only the final click. Record when people hesitate, search elsewhere, ask for help, return to prior text or invent a rule. Distinguish interface affordance failures from linguistic ones by testing a clearer layout or comparing against an approved source-language version where relevant.

7. Use contrastive probes to reveal missing conditions

A message often looks fine until a boundary case is introduced. If a refund applies only before dispatch, ask what happens one minute after dispatch. If a consent choice covers analytics but not account security, ask which feature continues after opting out. A reader who answers both examples identically may have flattened an essential condition.

Contrastive tasks work best when one small variable changes at a time. Hold wording, actor and context constant while varying a deadline, exception or state. Do not use a trick question that depends on unrelated world knowledge. The purpose is to expose an interpretation rule, not to defeat the participant.

Record the variable that changes and the inference that should change with it. If results are ambiguous, revise the test rather than forcing a pass or fail. A successful translation supports the intended rule across reasonable neighboring cases, not only the particular example chosen by its author.

8. Treat numbers and time as meaning, not typography

Amounts, dates, units, thresholds and durations can be displayed correctly yet misinterpreted. A deadline “by the end of the next business day” has an implicit counting convention. A notice “up to three attempts” may be read as three additional attempts after one failure. A range may look like an exact value if qualifiers disappear in translation.

Write the counting rule or uncertainty as an interpretation invariant. Then use a tiny set of test inputs straddling the boundary: just before, exactly at and just after the threshold. If an interface formats a date differently in different languages, test the value as the reader sees it and do not merely inspect the underlying timestamp.

Do not invent conversion rules during translation. Where a policy defines its own time or rounding method, use that approved rule. Escalate genuinely ambiguous source text rather than creating a locale-specific rule that did not exist.

9. Understand actors, reference and responsibility

Pronouns, grammatical subjects, passive voice and omitted agents can obscure who must act. “They will notify you when it is approved” may leave open whether approval is automatic or requires an administrator. A translator can make the sentence elegant while accidentally assigning responsibility to a different party.

Create an actor map for every consequential verb: who requests, who reviews, who authorizes, who receives information, who may appeal and who has permission to reverse an action. A participant can then reconstruct the chain in their own words. Ask “Who does what?” rather than “Do you understand who approves it?”

Some languages routinely omit subjects when context makes them recoverable. The goal is not to impose English syntax. The goal is to ensure that the intended actor remains recoverable in the actual setting, including when a sentence appears alone in a notification.

10. Handle culturally plausible but incorrect inferences

Words live in institutional practices. A “deposit” may imply refundable security money in one service and an advance payment in another. “Verification” may imply identity confirmation rather than final approval. When translators pick familiar labels, readers may import an entire procedure that the source never promised.

Do not classify every culturally different expectation as a translation error. Record what the reader expected, where that expectation came from, and whether the source or interface communicates the policy sufficiently. Sometimes the best fix is a short qualifying sentence rather than a new dictionary synonym.

Test the interpretation as a practical consequence. Ask whether the money is returned, whether the status is final, or whether a person may proceed. A target-language version is useful when it makes the actual institution legible, not when it merely sounds like a familiar institution.

11. Sampling and evidence without false certainty

A small qualitative comprehension study can reveal failure modes quickly, but it cannot estimate every reader’s probability of success without an appropriate sampling design and sufficient data. Participant recruitment should reflect the risk profile and the major kinds of intended readers. A single expert who knows the source process is not an adequate stand-in for a novice.

Predeclare which types of misunderstanding would stop release. Record the task, participant characteristics relevant to the task, exact prompt, response, observed action, suspected cause and revision. Do not invent sample-size rules or numerical acceptance thresholds from an unrelated published standard.

For a high-consequence message, use specialist review, representative user work and independent evidence together. If exposure is limited or recruitment difficult, report the limitation. A good report is actionable precisely because it does not overstate what a small study proves.

12. Error attribution and the repair hierarchy

When a test fails, resist reflexively blaming the translator. The source may be logically ambiguous. The translation may alter a quantifier. The interface may hide an exception. A reader may lack domain vocabulary. A button may carry a misleading label even when the paragraph is perfect. Each category calls for a different intervention.

Begin with the evidence: what did the participant believe, what text did they see, and what would they have done? Compare their interpretation with the frozen interpretation contract. Trace the failure to its smallest credible cause. Then change one element, preserve all other requirements and test again.

Maintain a decision record: source clarification, target wording change, design correction, supplemental example or training support. A repair is not complete when the editors prefer the new sentence; it is complete when new evidence shows the original failure mechanism no longer appears under comparable conditions.

Part II. Repeatable protocol: from intake to release

Protocol 1: Intake and risk

Write the user task as a verb and specify why failure matters. Identify whether the message is descriptive, transactional, contractual, instructional or urgent. Risk classification should be grounded in the actual consequences, not a blanket claim that every confusing word is dangerous.

Output: an interpretation contract and a named domain owner. Decision gate: a missing output means the next step has no stable evidence base. Preserve the material, its version, the exact test prompt, and the reason an interpretation was marked correct or incorrect. This prevents an editor from silently changing the acceptance criterion after a participant discovers a surprising reading.

Protocol 2: Source clarification

Read the source before commissioning another language version. Flag unstated counting rules, ambiguous actors, unresolved policy exceptions and inconsistent vocabulary across adjacent screens. Ask the owner to resolve ambiguity in the source instead of allowing each locale to improvise.

Output: an approved source baseline and an uncertainty log. Decision gate: a missing output means the next step has no stable evidence base. Preserve the material, its version, the exact test prompt, and the reason an interpretation was marked correct or incorrect. This prevents an editor from silently changing the acceptance criterion after a participant discovers a surprising reading.

Protocol 3: Translation context

Give the linguist the whole task flow and representative screenshots. Include previous and next states, labels, error messages, automated emails and the user’s expected knowledge. A string exported without context is not a sufficient comprehension fixture.

Output: a target artifact in realistic surroundings. Decision gate: a missing output means the next step has no stable evidence base. Preserve the material, its version, the exact test prompt, and the reason an interpretation was marked correct or incorrect. This prevents an editor from silently changing the acceptance criterion after a participant discovers a surprising reading.

Protocol 4: Recruitment

Identify novice and experienced readers, device access patterns, accessibility needs, education or literacy constraints and language variety relevant to the content. Recruit without treating national origin as a proxy for proficiency or comprehension.

Output: inclusion criteria, screening rationale and documented limits. Decision gate: a missing output means the next step has no stable evidence base. Preserve the material, its version, the exact test prompt, and the reason an interpretation was marked correct or incorrect. This prevents an editor from silently changing the acceptance criterion after a participant discovers a surprising reading.

Protocol 5: Moderator briefing

Teach moderators the invariant facts and common wrong inferences without inviting them to lead respondents. Preserve exact source of prompts and define when to stop or redirect a participant. Conduct a practice interview to discover suggestive wording.

Output: a stable, neutral interview guide. Decision gate: a missing output means the next step has no stable evidence base. Preserve the material, its version, the exact test prompt, and the reason an interpretation was marked correct or incorrect. This prevents an editor from silently changing the acceptance criterion after a participant discovers a surprising reading.

Protocol 6: Baseline exposure

Show the participant the artifact in the same medium and state as a real encounter. No prior vocabulary lesson, glossary explanation or facilitator summary should precede the first observation unless real users would receive one.

Output: first-exposure actions, timing and spontaneous comments. Decision gate: a missing output means the next step has no stable evidence base. Preserve the material, its version, the exact test prompt, and the reason an interpretation was marked correct or incorrect. This prevents an editor from silently changing the acceptance criterion after a participant discovers a surprising reading.

Protocol 7: Paraphrase

Ask the reader to explain the message in ordinary terms. Record who acts, the timing, the qualification and the consequence. Distinguish someone who merely repeats a phrase from someone who can predict its practical effect.

Output: a reader interpretation map compared against the invariant. Decision gate: a missing output means the next step has no stable evidence base. Preserve the material, its version, the exact test prompt, and the reason an interpretation was marked correct or incorrect. This prevents an editor from silently changing the acceptance criterion after a participant discovers a surprising reading.

Protocol 8: Scenario task

Give the reader a realistic fictional problem and let them select an action. Include at least one plausible wrong alternative. Note where they search for evidence, whether they change their mind, and whether the UI induces the mistake.

Output: trace of action, reason and supporting evidence. Decision gate: a missing output means the next step has no stable evidence base. Preserve the material, its version, the exact test prompt, and the reason an interpretation was marked correct or incorrect. This prevents an editor from silently changing the acceptance criterion after a participant discovers a surprising reading.

Protocol 9: Boundary testing

Change one relevant detail: a deadline, quantity, actor, permission, status or exception. Ask the same decision question. Keep the context otherwise constant so a shift in answer reveals the rule being used.

Output: a contrastive pair with an identified semantic hinge. Decision gate: a missing output means the next step has no stable evidence base. Preserve the material, its version, the exact test prompt, and the reason an interpretation was marked correct or incorrect. This prevents an editor from silently changing the acceptance criterion after a participant discovers a surprising reading.

Protocol 10: Cause classification

Compare the participant’s answer with the contract and inspect source, translation, interaction design, formatting and background knowledge. The category may be mixed: an obscure term and a hidden footnote may reinforce each other.

Output: an evidence-based failure classification and severity. Decision gate: a missing output means the next step has no stable evidence base. Preserve the material, its version, the exact test prompt, and the reason an interpretation was marked correct or incorrect. This prevents an editor from silently changing the acceptance criterion after a participant discovers a surprising reading.

Protocol 11: Repair and review

Make the narrowest change that communicates the intended rule, then obtain the appropriate linguistic and domain sign-offs. Where a source must change, re-evaluate all language versions. Record why alternatives were rejected.

Output: a versioned revision and owner decision. Decision gate: a missing output means the next step has no stable evidence base. Preserve the material, its version, the exact test prompt, and the reason an interpretation was marked correct or incorrect. This prevents an editor from silently changing the acceptance criterion after a participant discovers a surprising reading.

Protocol 12: Independent retest

Test the revision under comparable conditions with participants who have not been coached into the answer. Keep pre- and post-revision evidence separate. Report remaining limitations and do not equate a handful of successes with population-wide proof.

Output: a release recommendation with explicit residual uncertainty. Decision gate: a missing output means the next step has no stable evidence base. Preserve the material, its version, the exact test prompt, and the reason an interpretation was marked correct or incorrect. This prevents an editor from silently changing the acceptance criterion after a participant discovers a surprising reading.

Part III. Field laboratories: testing meaning in 24 different reader tasks

The following laboratories use fictional examples constructed to teach diagnostic reasoning. They are not user research findings, official instructions, translations supplied by a professional or approved policies. Each laboratory changes the test method while keeping a concrete source intention visible. The repeated structure is deliberate: it lets a reviewer compare several evidence types for the same meaning risk, while the observed variables and decisions are different.

Domain studio 1: Subscription renewal notice

A a subscriber viewing a billing reminder encounters a cancellation control shown beside a renewal date. The intended rule is: The plan renews on the displayed date unless cancellation is submitted beforehand; access continues until the end of the paid period. The risky alternative is cancellation immediately terminates an already paid service. This matters because an unexpected charge or mistaken loss of access. Here the semantic hinge is that renewal prevention and continued access are separate outcomes. These are invented teaching fixtures, so a real publisher must replace them with an approved source contract and its own audience research.

Diagnostic case 001: Subscription renewal notice × Blind paraphrase

Reader-facing test. For a subscriber viewing a billing reminder, begin with a cancellation control shown beside a renewal date and do not disclose the intended interpretation. Withhold the source original and ask the reader, in their own language, to explain the message as if describing it to someone who has not seen the screen. The working question is whether the person can distinguish “renewal prevention and continued access are separate outcomes” from the tempting claim that cancellation immediately terminates an already paid service. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Listen for the causal story: what triggers the action, what stays true afterward and which actor controls the change. The expected observation is that the person distinguishes the charge deadline from the end of current access. If instead the answer implies that cancellation immediately terminates an already paid service, log the exact words and where the reader looked. The potential harm is an unexpected charge or mistaken loss of access. Do not treat a fluent reformulation as proof of meaning; the word ‘cancel’ carries both a billing and a service-state meaning. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to show the renewal condition and paid-through date together. A clean paraphrase should preserve the decision rule rather than repeat isolated keywords. A polished but incorrect summary is more informative than a low-confidence guess because it shows the exact rule the reader has constructed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 002: Subscription renewal notice × Task simulation

Reader-facing test. For a subscriber viewing a billing reminder, begin with a cancellation control shown beside a renewal date and do not disclose the intended interpretation. Give a fictional scenario with a choice between acting now, waiting for confirmation and searching for another condition; allow the reader to use only the visible translated material. The working question is whether the person can distinguish “renewal prevention and continued access are separate outcomes” from the tempting claim that cancellation immediately terminates an already paid service. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Capture the chosen action, the text used to justify it and whether a wrong branch seems attractive. The expected observation is that the person distinguishes the charge deadline from the end of current access. If instead the answer implies that cancellation immediately terminates an already paid service, log the exact words and where the reader looked. The potential harm is an unexpected charge or mistaken loss of access. Do not treat a fluent reformulation as proof of meaning; the word ‘cancel’ carries both a billing and a service-state meaning. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to show the renewal condition and paid-through date together. Do not score the final click alone. Someone may guess correctly from placement while holding the wrong belief. A complete result includes the reason, expected consequence and what additional information the person believes is needed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 003: Subscription renewal notice × Think-aloud trace

Reader-facing test. For a subscriber viewing a billing reminder, begin with a cancellation control shown beside a renewal date and do not disclose the intended interpretation. Ask the reader to narrate thoughts while moving from the notice to the relevant action, but do not correct terminology or explain the organization’s policy. The working question is whether the person can distinguish “renewal prevention and continued access are separate outcomes” from the tempting claim that cancellation immediately terminates an already paid service. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Mark the moment a term is interpreted, the moment confidence changes and the point at which layout or icons redirect attention. The expected observation is that the person distinguishes the charge deadline from the end of current access. If instead the answer implies that cancellation immediately terminates an already paid service, log the exact words and where the reader looked. The potential harm is an unexpected charge or mistaken loss of access. Do not treat a fluent reformulation as proof of meaning; the word ‘cancel’ carries both a billing and a service-state meaning. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to show the renewal condition and paid-through date together. This method can identify an interface cue rather than a translation error. The very act of thinking aloud may change behavior, so report it as qualitative diagnostic evidence rather than a naturalistic time-to-completion measurement. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 004: Subscription renewal notice × Contrastive boundary pair

Reader-facing test. For a subscriber viewing a billing reminder, begin with a cancellation control shown beside a renewal date and do not disclose the intended interpretation. Present two nearly identical fictional cases that differ only in the decisive condition; ask whether the outcome changes and why. The working question is whether the person can distinguish “renewal prevention and continued access are separate outcomes” from the tempting claim that cancellation immediately terminates an already paid service. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Look for the reader’s treatment of a status, exception or timing boundary that the source intended to preserve. The expected observation is that the person distinguishes the charge deadline from the end of current access. If instead the answer implies that cancellation immediately terminates an already paid service, log the exact words and where the reader looked. The potential harm is an unexpected charge or mistaken loss of access. Do not treat a fluent reformulation as proof of meaning; the word ‘cancel’ carries both a billing and a service-state meaning. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to show the renewal condition and paid-through date together. A reader who gives the same answer to both cases may have lost the qualifier. However, if the source rule itself is ambiguous, pause the test and clarify it with the content owner. Do not force respondents to guess at a rule the author has not specified. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 005: Subscription renewal notice × Actor and responsibility map

Reader-facing test. For a subscriber viewing a billing reminder, begin with a cancellation control shown beside a renewal date and do not disclose the intended interpretation. Ask the reader to identify who submits, who decides, who can revise the decision and who is affected; let them point to the translated words supporting each answer. The working question is whether the person can distinguish “renewal prevention and continued access are separate outcomes” from the tempting claim that cancellation immediately terminates an already paid service. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare the named actors with the interpretation contract and note any responsibility the reader invents. The expected observation is that the person distinguishes the charge deadline from the end of current access. If instead the answer implies that cancellation immediately terminates an already paid service, log the exact words and where the reader looked. The potential harm is an unexpected charge or mistaken loss of access. Do not treat a fluent reformulation as proof of meaning; the word ‘cancel’ carries both a billing and a service-state meaning. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to show the renewal condition and paid-through date together. Passive voice is not automatically wrong, and some languages omit subjects naturally. The test is whether the actual actor is recoverable. If the reader supplies an unintended actor, identify whether nearby labels, prior notifications or grammatical choices created that inference. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 006: Subscription renewal notice × Counterexample challenge

Reader-facing test. For a subscriber viewing a billing reminder, begin with a cancellation control shown beside a renewal date and do not disclose the intended interpretation. Invite the reader to invent an example in which the notice would not apply, or to say what evidence would make the current instruction no longer valid. The working question is whether the person can distinguish “renewal prevention and continued access are separate outcomes” from the tempting claim that cancellation immediately terminates an already paid service. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Inspect whether the reader recognizes limits, exceptions, unknown information and provisional states. The expected observation is that the person distinguishes the charge deadline from the end of current access. If instead the answer implies that cancellation immediately terminates an already paid service, log the exact words and where the reader looked. The potential harm is an unexpected charge or mistaken loss of access. Do not treat a fluent reformulation as proof of meaning; the word ‘cancel’ carries both a billing and a service-state meaning. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to show the renewal condition and paid-through date together. An inability to imagine a counterexample may be a testing skill issue, so offer a neutral hypothetical rather than marking immediate failure. The key evidence is the reader’s understanding of the boundary after an ordinary scenario is supplied without revealing the answer. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 007: Subscription renewal notice × Delayed teach-back

Reader-facing test. For a subscriber viewing a billing reminder, begin with a cancellation control shown beside a renewal date and do not disclose the intended interpretation. After an unrelated short filler activity, ask the reader to explain the next step to a fictional colleague who cannot view the translated message. The working question is whether the person can distinguish “renewal prevention and continued access are separate outcomes” from the tempting claim that cancellation immediately terminates an already paid service. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Note which conditions survive memory, which qualifiers disappear and whether confidence rises as detail is lost. The expected observation is that the person distinguishes the charge deadline from the end of current access. If instead the answer implies that cancellation immediately terminates an already paid service, log the exact words and where the reader looked. The potential harm is an unexpected charge or mistaken loss of access. Do not treat a fluent reformulation as proof of meaning; the word ‘cancel’ carries both a billing and a service-state meaning. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to show the renewal condition and paid-through date together. Teach-back tests retained practical meaning, not memorization of exact wording. Compare it with first-exposure paraphrase. If the reader initially understood but later drops a critical exception, consider information placement, a reminder, a concise summary or a clearer call to action. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 008: Subscription renewal notice × Alternative wording comparison

Reader-facing test. For a subscriber viewing a billing reminder, begin with a cancellation control shown beside a renewal date and do not disclose the intended interpretation. Show two approved candidate renderings in a balanced order across participants and ask what each would lead them to do; avoid asking which sounds more elegant. The working question is whether the person can distinguish “renewal prevention and continued access are separate outcomes” from the tempting claim that cancellation immediately terminates an already paid service. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Record whether the alternatives produce different real-world predictions despite similar fluency judgments. The expected observation is that the person distinguishes the charge deadline from the end of current access. If instead the answer implies that cancellation immediately terminates an already paid service, log the exact words and where the reader looked. The potential harm is an unexpected charge or mistaken loss of access. Do not treat a fluent reformulation as proof of meaning; the word ‘cancel’ carries both a billing and a service-state meaning. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to show the renewal condition and paid-through date together. Do not optimize language by majority preference alone: a popular label can still induce a serious wrong decision. Preference is a supporting measure. Select wording that better preserves the rule, is natural for the intended audience and survives a subsequent task-based retest. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 009: Subscription renewal notice × Release-readiness retest

Reader-facing test. For a subscriber viewing a billing reminder, begin with a cancellation control shown beside a renewal date and do not disclose the intended interpretation. Present the repaired artifact with the same surrounding interface but an independent fictional case; recruit someone not coached by the earlier test where feasible. The working question is whether the person can distinguish “renewal prevention and continued access are separate outcomes” from the tempting claim that cancellation immediately terminates an already paid service. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare unprompted decisions, explanations and remaining errors rather than merely tallying favorable reactions. The expected observation is that the person distinguishes the charge deadline from the end of current access. If instead the answer implies that cancellation immediately terminates an already paid service, log the exact words and where the reader looked. The potential harm is an unexpected charge or mistaken loss of access. Do not treat a fluent reformulation as proof of meaning; the word ‘cancel’ carries both a billing and a service-state meaning. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to show the renewal condition and paid-through date together. A corrected phrase is not validated by the editor’s confidence. Preserve before-and-after versions, source approval, exact prompts, reader evidence and unresolved risks. Report that the tested failure diminished under these conditions, not that the translation is universally impossible to misunderstand. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Domain studio 2: Shared-device sign-out

A a reader leaving a library or borrowed device encounters the ‘Sign out’ button and a nearby download list. The intended rule is: Signing out ends this browser session, but previously downloaded files may remain on the device until separately removed. The risky alternative is signing out deletes every locally stored document. This matters because private files remain accessible to the next device user. Here the semantic hinge is that session authentication differs from local file persistence. These are invented teaching fixtures, so a real publisher must replace them with an approved source contract and its own audience research.

Diagnostic case 010: Shared-device sign-out × Blind paraphrase

Reader-facing test. For a reader leaving a library or borrowed device, begin with the ‘Sign out’ button and a nearby download list and do not disclose the intended interpretation. Withhold the source original and ask the reader, in their own language, to explain the message as if describing it to someone who has not seen the screen. The working question is whether the person can distinguish “session authentication differs from local file persistence” from the tempting claim that signing out deletes every locally stored document. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Listen for the causal story: what triggers the action, what stays true afterward and which actor controls the change. The expected observation is that the person can identify the separate clean-up action without coaching. If instead the answer implies that signing out deletes every locally stored document, log the exact words and where the reader looked. The potential harm is private files remain accessible to the next device user. Do not treat a fluent reformulation as proof of meaning; a security verb does not describe every storage state. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to state the remaining-file risk beside the sign-out action. A clean paraphrase should preserve the decision rule rather than repeat isolated keywords. A polished but incorrect summary is more informative than a low-confidence guess because it shows the exact rule the reader has constructed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 011: Shared-device sign-out × Task simulation

Reader-facing test. For a reader leaving a library or borrowed device, begin with the ‘Sign out’ button and a nearby download list and do not disclose the intended interpretation. Give a fictional scenario with a choice between acting now, waiting for confirmation and searching for another condition; allow the reader to use only the visible translated material. The working question is whether the person can distinguish “session authentication differs from local file persistence” from the tempting claim that signing out deletes every locally stored document. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Capture the chosen action, the text used to justify it and whether a wrong branch seems attractive. The expected observation is that the person can identify the separate clean-up action without coaching. If instead the answer implies that signing out deletes every locally stored document, log the exact words and where the reader looked. The potential harm is private files remain accessible to the next device user. Do not treat a fluent reformulation as proof of meaning; a security verb does not describe every storage state. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to state the remaining-file risk beside the sign-out action. Do not score the final click alone. Someone may guess correctly from placement while holding the wrong belief. A complete result includes the reason, expected consequence and what additional information the person believes is needed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 012: Shared-device sign-out × Think-aloud trace

Reader-facing test. For a reader leaving a library or borrowed device, begin with the ‘Sign out’ button and a nearby download list and do not disclose the intended interpretation. Ask the reader to narrate thoughts while moving from the notice to the relevant action, but do not correct terminology or explain the organization’s policy. The working question is whether the person can distinguish “session authentication differs from local file persistence” from the tempting claim that signing out deletes every locally stored document. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Mark the moment a term is interpreted, the moment confidence changes and the point at which layout or icons redirect attention. The expected observation is that the person can identify the separate clean-up action without coaching. If instead the answer implies that signing out deletes every locally stored document, log the exact words and where the reader looked. The potential harm is private files remain accessible to the next device user. Do not treat a fluent reformulation as proof of meaning; a security verb does not describe every storage state. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to state the remaining-file risk beside the sign-out action. This method can identify an interface cue rather than a translation error. The very act of thinking aloud may change behavior, so report it as qualitative diagnostic evidence rather than a naturalistic time-to-completion measurement. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 013: Shared-device sign-out × Contrastive boundary pair

Reader-facing test. For a reader leaving a library or borrowed device, begin with the ‘Sign out’ button and a nearby download list and do not disclose the intended interpretation. Present two nearly identical fictional cases that differ only in the decisive condition; ask whether the outcome changes and why. The working question is whether the person can distinguish “session authentication differs from local file persistence” from the tempting claim that signing out deletes every locally stored document. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Look for the reader’s treatment of a status, exception or timing boundary that the source intended to preserve. The expected observation is that the person can identify the separate clean-up action without coaching. If instead the answer implies that signing out deletes every locally stored document, log the exact words and where the reader looked. The potential harm is private files remain accessible to the next device user. Do not treat a fluent reformulation as proof of meaning; a security verb does not describe every storage state. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to state the remaining-file risk beside the sign-out action. A reader who gives the same answer to both cases may have lost the qualifier. However, if the source rule itself is ambiguous, pause the test and clarify it with the content owner. Do not force respondents to guess at a rule the author has not specified. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 014: Shared-device sign-out × Actor and responsibility map

Reader-facing test. For a reader leaving a library or borrowed device, begin with the ‘Sign out’ button and a nearby download list and do not disclose the intended interpretation. Ask the reader to identify who submits, who decides, who can revise the decision and who is affected; let them point to the translated words supporting each answer. The working question is whether the person can distinguish “session authentication differs from local file persistence” from the tempting claim that signing out deletes every locally stored document. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare the named actors with the interpretation contract and note any responsibility the reader invents. The expected observation is that the person can identify the separate clean-up action without coaching. If instead the answer implies that signing out deletes every locally stored document, log the exact words and where the reader looked. The potential harm is private files remain accessible to the next device user. Do not treat a fluent reformulation as proof of meaning; a security verb does not describe every storage state. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to state the remaining-file risk beside the sign-out action. Passive voice is not automatically wrong, and some languages omit subjects naturally. The test is whether the actual actor is recoverable. If the reader supplies an unintended actor, identify whether nearby labels, prior notifications or grammatical choices created that inference. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 015: Shared-device sign-out × Counterexample challenge

Reader-facing test. For a reader leaving a library or borrowed device, begin with the ‘Sign out’ button and a nearby download list and do not disclose the intended interpretation. Invite the reader to invent an example in which the notice would not apply, or to say what evidence would make the current instruction no longer valid. The working question is whether the person can distinguish “session authentication differs from local file persistence” from the tempting claim that signing out deletes every locally stored document. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Inspect whether the reader recognizes limits, exceptions, unknown information and provisional states. The expected observation is that the person can identify the separate clean-up action without coaching. If instead the answer implies that signing out deletes every locally stored document, log the exact words and where the reader looked. The potential harm is private files remain accessible to the next device user. Do not treat a fluent reformulation as proof of meaning; a security verb does not describe every storage state. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to state the remaining-file risk beside the sign-out action. An inability to imagine a counterexample may be a testing skill issue, so offer a neutral hypothetical rather than marking immediate failure. The key evidence is the reader’s understanding of the boundary after an ordinary scenario is supplied without revealing the answer. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 016: Shared-device sign-out × Delayed teach-back

Reader-facing test. For a reader leaving a library or borrowed device, begin with the ‘Sign out’ button and a nearby download list and do not disclose the intended interpretation. After an unrelated short filler activity, ask the reader to explain the next step to a fictional colleague who cannot view the translated message. The working question is whether the person can distinguish “session authentication differs from local file persistence” from the tempting claim that signing out deletes every locally stored document. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Note which conditions survive memory, which qualifiers disappear and whether confidence rises as detail is lost. The expected observation is that the person can identify the separate clean-up action without coaching. If instead the answer implies that signing out deletes every locally stored document, log the exact words and where the reader looked. The potential harm is private files remain accessible to the next device user. Do not treat a fluent reformulation as proof of meaning; a security verb does not describe every storage state. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to state the remaining-file risk beside the sign-out action. Teach-back tests retained practical meaning, not memorization of exact wording. Compare it with first-exposure paraphrase. If the reader initially understood but later drops a critical exception, consider information placement, a reminder, a concise summary or a clearer call to action. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 017: Shared-device sign-out × Alternative wording comparison

Reader-facing test. For a reader leaving a library or borrowed device, begin with the ‘Sign out’ button and a nearby download list and do not disclose the intended interpretation. Show two approved candidate renderings in a balanced order across participants and ask what each would lead them to do; avoid asking which sounds more elegant. The working question is whether the person can distinguish “session authentication differs from local file persistence” from the tempting claim that signing out deletes every locally stored document. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Record whether the alternatives produce different real-world predictions despite similar fluency judgments. The expected observation is that the person can identify the separate clean-up action without coaching. If instead the answer implies that signing out deletes every locally stored document, log the exact words and where the reader looked. The potential harm is private files remain accessible to the next device user. Do not treat a fluent reformulation as proof of meaning; a security verb does not describe every storage state. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to state the remaining-file risk beside the sign-out action. Do not optimize language by majority preference alone: a popular label can still induce a serious wrong decision. Preference is a supporting measure. Select wording that better preserves the rule, is natural for the intended audience and survives a subsequent task-based retest. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 018: Shared-device sign-out × Release-readiness retest

Reader-facing test. For a reader leaving a library or borrowed device, begin with the ‘Sign out’ button and a nearby download list and do not disclose the intended interpretation. Present the repaired artifact with the same surrounding interface but an independent fictional case; recruit someone not coached by the earlier test where feasible. The working question is whether the person can distinguish “session authentication differs from local file persistence” from the tempting claim that signing out deletes every locally stored document. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare unprompted decisions, explanations and remaining errors rather than merely tallying favorable reactions. The expected observation is that the person can identify the separate clean-up action without coaching. If instead the answer implies that signing out deletes every locally stored document, log the exact words and where the reader looked. The potential harm is private files remain accessible to the next device user. Do not treat a fluent reformulation as proof of meaning; a security verb does not describe every storage state. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to state the remaining-file risk beside the sign-out action. A corrected phrase is not validated by the editor’s confidence. Preserve before-and-after versions, source approval, exact prompts, reader evidence and unresolved risks. Report that the tested failure diminished under these conditions, not that the translation is universally impossible to misunderstand. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Domain studio 3: Cloud-file deletion notice

A a team member removing a shared document encounters a delete confirmation dialog with recovery information. The intended rule is: Deleting the item removes it from the active shared view, while a limited recovery window may still exist under the service policy. The risky alternative is deletion guarantees immediate destruction of all copies. This matters because false privacy assurances or accidental loss of recoverable work. Here the semantic hinge is that visibility removal and retention are not identical. These are invented teaching fixtures, so a real publisher must replace them with an approved source contract and its own audience research.

Diagnostic case 019: Cloud-file deletion notice × Blind paraphrase

Reader-facing test. For a team member removing a shared document, begin with a delete confirmation dialog with recovery information and do not disclose the intended interpretation. Withhold the source original and ask the reader, in their own language, to explain the message as if describing it to someone who has not seen the screen. The working question is whether the person can distinguish “visibility removal and retention are not identical” from the tempting claim that deletion guarantees immediate destruction of all copies. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Listen for the causal story: what triggers the action, what stays true afterward and which actor controls the change. The expected observation is that the person names what disappears immediately and what may remain. If instead the answer implies that deletion guarantees immediate destruction of all copies, log the exact words and where the reader looked. The potential harm is false privacy assurances or accidental loss of recoverable work. Do not treat a fluent reformulation as proof of meaning; destruction, hiding and recoverability are separate concepts. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to identify the exact state change and where the retention rule is stated. A clean paraphrase should preserve the decision rule rather than repeat isolated keywords. A polished but incorrect summary is more informative than a low-confidence guess because it shows the exact rule the reader has constructed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 020: Cloud-file deletion notice × Task simulation

Reader-facing test. For a team member removing a shared document, begin with a delete confirmation dialog with recovery information and do not disclose the intended interpretation. Give a fictional scenario with a choice between acting now, waiting for confirmation and searching for another condition; allow the reader to use only the visible translated material. The working question is whether the person can distinguish “visibility removal and retention are not identical” from the tempting claim that deletion guarantees immediate destruction of all copies. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Capture the chosen action, the text used to justify it and whether a wrong branch seems attractive. The expected observation is that the person names what disappears immediately and what may remain. If instead the answer implies that deletion guarantees immediate destruction of all copies, log the exact words and where the reader looked. The potential harm is false privacy assurances or accidental loss of recoverable work. Do not treat a fluent reformulation as proof of meaning; destruction, hiding and recoverability are separate concepts. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to identify the exact state change and where the retention rule is stated. Do not score the final click alone. Someone may guess correctly from placement while holding the wrong belief. A complete result includes the reason, expected consequence and what additional information the person believes is needed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 021: Cloud-file deletion notice × Think-aloud trace

Reader-facing test. For a team member removing a shared document, begin with a delete confirmation dialog with recovery information and do not disclose the intended interpretation. Ask the reader to narrate thoughts while moving from the notice to the relevant action, but do not correct terminology or explain the organization’s policy. The working question is whether the person can distinguish “visibility removal and retention are not identical” from the tempting claim that deletion guarantees immediate destruction of all copies. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Mark the moment a term is interpreted, the moment confidence changes and the point at which layout or icons redirect attention. The expected observation is that the person names what disappears immediately and what may remain. If instead the answer implies that deletion guarantees immediate destruction of all copies, log the exact words and where the reader looked. The potential harm is false privacy assurances or accidental loss of recoverable work. Do not treat a fluent reformulation as proof of meaning; destruction, hiding and recoverability are separate concepts. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to identify the exact state change and where the retention rule is stated. This method can identify an interface cue rather than a translation error. The very act of thinking aloud may change behavior, so report it as qualitative diagnostic evidence rather than a naturalistic time-to-completion measurement. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 022: Cloud-file deletion notice × Contrastive boundary pair

Reader-facing test. For a team member removing a shared document, begin with a delete confirmation dialog with recovery information and do not disclose the intended interpretation. Present two nearly identical fictional cases that differ only in the decisive condition; ask whether the outcome changes and why. The working question is whether the person can distinguish “visibility removal and retention are not identical” from the tempting claim that deletion guarantees immediate destruction of all copies. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Look for the reader’s treatment of a status, exception or timing boundary that the source intended to preserve. The expected observation is that the person names what disappears immediately and what may remain. If instead the answer implies that deletion guarantees immediate destruction of all copies, log the exact words and where the reader looked. The potential harm is false privacy assurances or accidental loss of recoverable work. Do not treat a fluent reformulation as proof of meaning; destruction, hiding and recoverability are separate concepts. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to identify the exact state change and where the retention rule is stated. A reader who gives the same answer to both cases may have lost the qualifier. However, if the source rule itself is ambiguous, pause the test and clarify it with the content owner. Do not force respondents to guess at a rule the author has not specified. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 023: Cloud-file deletion notice × Actor and responsibility map

Reader-facing test. For a team member removing a shared document, begin with a delete confirmation dialog with recovery information and do not disclose the intended interpretation. Ask the reader to identify who submits, who decides, who can revise the decision and who is affected; let them point to the translated words supporting each answer. The working question is whether the person can distinguish “visibility removal and retention are not identical” from the tempting claim that deletion guarantees immediate destruction of all copies. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare the named actors with the interpretation contract and note any responsibility the reader invents. The expected observation is that the person names what disappears immediately and what may remain. If instead the answer implies that deletion guarantees immediate destruction of all copies, log the exact words and where the reader looked. The potential harm is false privacy assurances or accidental loss of recoverable work. Do not treat a fluent reformulation as proof of meaning; destruction, hiding and recoverability are separate concepts. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to identify the exact state change and where the retention rule is stated. Passive voice is not automatically wrong, and some languages omit subjects naturally. The test is whether the actual actor is recoverable. If the reader supplies an unintended actor, identify whether nearby labels, prior notifications or grammatical choices created that inference. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 024: Cloud-file deletion notice × Counterexample challenge

Reader-facing test. For a team member removing a shared document, begin with a delete confirmation dialog with recovery information and do not disclose the intended interpretation. Invite the reader to invent an example in which the notice would not apply, or to say what evidence would make the current instruction no longer valid. The working question is whether the person can distinguish “visibility removal and retention are not identical” from the tempting claim that deletion guarantees immediate destruction of all copies. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Inspect whether the reader recognizes limits, exceptions, unknown information and provisional states. The expected observation is that the person names what disappears immediately and what may remain. If instead the answer implies that deletion guarantees immediate destruction of all copies, log the exact words and where the reader looked. The potential harm is false privacy assurances or accidental loss of recoverable work. Do not treat a fluent reformulation as proof of meaning; destruction, hiding and recoverability are separate concepts. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to identify the exact state change and where the retention rule is stated. An inability to imagine a counterexample may be a testing skill issue, so offer a neutral hypothetical rather than marking immediate failure. The key evidence is the reader’s understanding of the boundary after an ordinary scenario is supplied without revealing the answer. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 025: Cloud-file deletion notice × Delayed teach-back

Reader-facing test. For a team member removing a shared document, begin with a delete confirmation dialog with recovery information and do not disclose the intended interpretation. After an unrelated short filler activity, ask the reader to explain the next step to a fictional colleague who cannot view the translated message. The working question is whether the person can distinguish “visibility removal and retention are not identical” from the tempting claim that deletion guarantees immediate destruction of all copies. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Note which conditions survive memory, which qualifiers disappear and whether confidence rises as detail is lost. The expected observation is that the person names what disappears immediately and what may remain. If instead the answer implies that deletion guarantees immediate destruction of all copies, log the exact words and where the reader looked. The potential harm is false privacy assurances or accidental loss of recoverable work. Do not treat a fluent reformulation as proof of meaning; destruction, hiding and recoverability are separate concepts. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to identify the exact state change and where the retention rule is stated. Teach-back tests retained practical meaning, not memorization of exact wording. Compare it with first-exposure paraphrase. If the reader initially understood but later drops a critical exception, consider information placement, a reminder, a concise summary or a clearer call to action. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 026: Cloud-file deletion notice × Alternative wording comparison

Reader-facing test. For a team member removing a shared document, begin with a delete confirmation dialog with recovery information and do not disclose the intended interpretation. Show two approved candidate renderings in a balanced order across participants and ask what each would lead them to do; avoid asking which sounds more elegant. The working question is whether the person can distinguish “visibility removal and retention are not identical” from the tempting claim that deletion guarantees immediate destruction of all copies. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Record whether the alternatives produce different real-world predictions despite similar fluency judgments. The expected observation is that the person names what disappears immediately and what may remain. If instead the answer implies that deletion guarantees immediate destruction of all copies, log the exact words and where the reader looked. The potential harm is false privacy assurances or accidental loss of recoverable work. Do not treat a fluent reformulation as proof of meaning; destruction, hiding and recoverability are separate concepts. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to identify the exact state change and where the retention rule is stated. Do not optimize language by majority preference alone: a popular label can still induce a serious wrong decision. Preference is a supporting measure. Select wording that better preserves the rule, is natural for the intended audience and survives a subsequent task-based retest. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 027: Cloud-file deletion notice × Release-readiness retest

Reader-facing test. For a team member removing a shared document, begin with a delete confirmation dialog with recovery information and do not disclose the intended interpretation. Present the repaired artifact with the same surrounding interface but an independent fictional case; recruit someone not coached by the earlier test where feasible. The working question is whether the person can distinguish “visibility removal and retention are not identical” from the tempting claim that deletion guarantees immediate destruction of all copies. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare unprompted decisions, explanations and remaining errors rather than merely tallying favorable reactions. The expected observation is that the person names what disappears immediately and what may remain. If instead the answer implies that deletion guarantees immediate destruction of all copies, log the exact words and where the reader looked. The potential harm is false privacy assurances or accidental loss of recoverable work. Do not treat a fluent reformulation as proof of meaning; destruction, hiding and recoverability are separate concepts. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to identify the exact state change and where the retention rule is stated. A corrected phrase is not validated by the editor’s confidence. Preserve before-and-after versions, source approval, exact prompts, reader evidence and unresolved risks. Report that the tested failure diminished under these conditions, not that the translation is universally impossible to misunderstand. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Domain studio 4: Returns and refund portal

A a purchaser looking at return conditions encounters a return button next to a status badge. The intended rule is: A request is eligible during the stated window, but payment is reimbursed only after the returned item passes the published checks. The risky alternative is request eligibility guarantees an immediate refund. This matters because unrealistic expectations and unnecessary disputes. Here the semantic hinge is that submission, inspection and reimbursement are distinct stages. These are invented teaching fixtures, so a real publisher must replace them with an approved source contract and its own audience research.

Diagnostic case 028: Returns and refund portal × Blind paraphrase

Reader-facing test. For a purchaser looking at return conditions, begin with a return button next to a status badge and do not disclose the intended interpretation. Withhold the source original and ask the reader, in their own language, to explain the message as if describing it to someone who has not seen the screen. The working question is whether the person can distinguish “submission, inspection and reimbursement are distinct stages” from the tempting claim that request eligibility guarantees an immediate refund. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Listen for the causal story: what triggers the action, what stays true afterward and which actor controls the change. The expected observation is that the person can predict which action starts the process and which event releases money. If instead the answer implies that request eligibility guarantees an immediate refund, log the exact words and where the reader looked. The potential harm is unrealistic expectations and unnecessary disputes. Do not treat a fluent reformulation as proof of meaning; eligibility is not a completed financial outcome. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label the request stage and approval stage separately. A clean paraphrase should preserve the decision rule rather than repeat isolated keywords. A polished but incorrect summary is more informative than a low-confidence guess because it shows the exact rule the reader has constructed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 029: Returns and refund portal × Task simulation

Reader-facing test. For a purchaser looking at return conditions, begin with a return button next to a status badge and do not disclose the intended interpretation. Give a fictional scenario with a choice between acting now, waiting for confirmation and searching for another condition; allow the reader to use only the visible translated material. The working question is whether the person can distinguish “submission, inspection and reimbursement are distinct stages” from the tempting claim that request eligibility guarantees an immediate refund. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Capture the chosen action, the text used to justify it and whether a wrong branch seems attractive. The expected observation is that the person can predict which action starts the process and which event releases money. If instead the answer implies that request eligibility guarantees an immediate refund, log the exact words and where the reader looked. The potential harm is unrealistic expectations and unnecessary disputes. Do not treat a fluent reformulation as proof of meaning; eligibility is not a completed financial outcome. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label the request stage and approval stage separately. Do not score the final click alone. Someone may guess correctly from placement while holding the wrong belief. A complete result includes the reason, expected consequence and what additional information the person believes is needed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 030: Returns and refund portal × Think-aloud trace

Reader-facing test. For a purchaser looking at return conditions, begin with a return button next to a status badge and do not disclose the intended interpretation. Ask the reader to narrate thoughts while moving from the notice to the relevant action, but do not correct terminology or explain the organization’s policy. The working question is whether the person can distinguish “submission, inspection and reimbursement are distinct stages” from the tempting claim that request eligibility guarantees an immediate refund. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Mark the moment a term is interpreted, the moment confidence changes and the point at which layout or icons redirect attention. The expected observation is that the person can predict which action starts the process and which event releases money. If instead the answer implies that request eligibility guarantees an immediate refund, log the exact words and where the reader looked. The potential harm is unrealistic expectations and unnecessary disputes. Do not treat a fluent reformulation as proof of meaning; eligibility is not a completed financial outcome. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label the request stage and approval stage separately. This method can identify an interface cue rather than a translation error. The very act of thinking aloud may change behavior, so report it as qualitative diagnostic evidence rather than a naturalistic time-to-completion measurement. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 031: Returns and refund portal × Contrastive boundary pair

Reader-facing test. For a purchaser looking at return conditions, begin with a return button next to a status badge and do not disclose the intended interpretation. Present two nearly identical fictional cases that differ only in the decisive condition; ask whether the outcome changes and why. The working question is whether the person can distinguish “submission, inspection and reimbursement are distinct stages” from the tempting claim that request eligibility guarantees an immediate refund. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Look for the reader’s treatment of a status, exception or timing boundary that the source intended to preserve. The expected observation is that the person can predict which action starts the process and which event releases money. If instead the answer implies that request eligibility guarantees an immediate refund, log the exact words and where the reader looked. The potential harm is unrealistic expectations and unnecessary disputes. Do not treat a fluent reformulation as proof of meaning; eligibility is not a completed financial outcome. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label the request stage and approval stage separately. A reader who gives the same answer to both cases may have lost the qualifier. However, if the source rule itself is ambiguous, pause the test and clarify it with the content owner. Do not force respondents to guess at a rule the author has not specified. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 032: Returns and refund portal × Actor and responsibility map

Reader-facing test. For a purchaser looking at return conditions, begin with a return button next to a status badge and do not disclose the intended interpretation. Ask the reader to identify who submits, who decides, who can revise the decision and who is affected; let them point to the translated words supporting each answer. The working question is whether the person can distinguish “submission, inspection and reimbursement are distinct stages” from the tempting claim that request eligibility guarantees an immediate refund. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare the named actors with the interpretation contract and note any responsibility the reader invents. The expected observation is that the person can predict which action starts the process and which event releases money. If instead the answer implies that request eligibility guarantees an immediate refund, log the exact words and where the reader looked. The potential harm is unrealistic expectations and unnecessary disputes. Do not treat a fluent reformulation as proof of meaning; eligibility is not a completed financial outcome. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label the request stage and approval stage separately. Passive voice is not automatically wrong, and some languages omit subjects naturally. The test is whether the actual actor is recoverable. If the reader supplies an unintended actor, identify whether nearby labels, prior notifications or grammatical choices created that inference. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 033: Returns and refund portal × Counterexample challenge

Reader-facing test. For a purchaser looking at return conditions, begin with a return button next to a status badge and do not disclose the intended interpretation. Invite the reader to invent an example in which the notice would not apply, or to say what evidence would make the current instruction no longer valid. The working question is whether the person can distinguish “submission, inspection and reimbursement are distinct stages” from the tempting claim that request eligibility guarantees an immediate refund. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Inspect whether the reader recognizes limits, exceptions, unknown information and provisional states. The expected observation is that the person can predict which action starts the process and which event releases money. If instead the answer implies that request eligibility guarantees an immediate refund, log the exact words and where the reader looked. The potential harm is unrealistic expectations and unnecessary disputes. Do not treat a fluent reformulation as proof of meaning; eligibility is not a completed financial outcome. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label the request stage and approval stage separately. An inability to imagine a counterexample may be a testing skill issue, so offer a neutral hypothetical rather than marking immediate failure. The key evidence is the reader’s understanding of the boundary after an ordinary scenario is supplied without revealing the answer. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 034: Returns and refund portal × Delayed teach-back

Reader-facing test. For a purchaser looking at return conditions, begin with a return button next to a status badge and do not disclose the intended interpretation. After an unrelated short filler activity, ask the reader to explain the next step to a fictional colleague who cannot view the translated message. The working question is whether the person can distinguish “submission, inspection and reimbursement are distinct stages” from the tempting claim that request eligibility guarantees an immediate refund. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Note which conditions survive memory, which qualifiers disappear and whether confidence rises as detail is lost. The expected observation is that the person can predict which action starts the process and which event releases money. If instead the answer implies that request eligibility guarantees an immediate refund, log the exact words and where the reader looked. The potential harm is unrealistic expectations and unnecessary disputes. Do not treat a fluent reformulation as proof of meaning; eligibility is not a completed financial outcome. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label the request stage and approval stage separately. Teach-back tests retained practical meaning, not memorization of exact wording. Compare it with first-exposure paraphrase. If the reader initially understood but later drops a critical exception, consider information placement, a reminder, a concise summary or a clearer call to action. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 035: Returns and refund portal × Alternative wording comparison

Reader-facing test. For a purchaser looking at return conditions, begin with a return button next to a status badge and do not disclose the intended interpretation. Show two approved candidate renderings in a balanced order across participants and ask what each would lead them to do; avoid asking which sounds more elegant. The working question is whether the person can distinguish “submission, inspection and reimbursement are distinct stages” from the tempting claim that request eligibility guarantees an immediate refund. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Record whether the alternatives produce different real-world predictions despite similar fluency judgments. The expected observation is that the person can predict which action starts the process and which event releases money. If instead the answer implies that request eligibility guarantees an immediate refund, log the exact words and where the reader looked. The potential harm is unrealistic expectations and unnecessary disputes. Do not treat a fluent reformulation as proof of meaning; eligibility is not a completed financial outcome. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label the request stage and approval stage separately. Do not optimize language by majority preference alone: a popular label can still induce a serious wrong decision. Preference is a supporting measure. Select wording that better preserves the rule, is natural for the intended audience and survives a subsequent task-based retest. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 036: Returns and refund portal × Release-readiness retest

Reader-facing test. For a purchaser looking at return conditions, begin with a return button next to a status badge and do not disclose the intended interpretation. Present the repaired artifact with the same surrounding interface but an independent fictional case; recruit someone not coached by the earlier test where feasible. The working question is whether the person can distinguish “submission, inspection and reimbursement are distinct stages” from the tempting claim that request eligibility guarantees an immediate refund. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare unprompted decisions, explanations and remaining errors rather than merely tallying favorable reactions. The expected observation is that the person can predict which action starts the process and which event releases money. If instead the answer implies that request eligibility guarantees an immediate refund, log the exact words and where the reader looked. The potential harm is unrealistic expectations and unnecessary disputes. Do not treat a fluent reformulation as proof of meaning; eligibility is not a completed financial outcome. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label the request stage and approval stage separately. A corrected phrase is not validated by the editor’s confidence. Preserve before-and-after versions, source approval, exact prompts, reader evidence and unresolved risks. Report that the tested failure diminished under these conditions, not that the translation is universally impossible to misunderstand. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Domain studio 5: Appointment preparation message

A a person checking nonclinical appointment logistics encounters a calendar card and preparation instructions. The intended rule is: The confirmation asks the visitor to arrive before the start time and bring an identity document; the arrival instruction does not authorize changes to treatment. The risky alternative is an early-arrival instruction is a different appointment time. This matters because a missed appointment or inappropriate assumptions about professional advice. Here the semantic hinge is that arrival buffer and scheduled start are distinct. These are invented teaching fixtures, so a real publisher must replace them with an approved source contract and its own audience research.

Diagnostic case 037: Appointment preparation message × Blind paraphrase

Reader-facing test. For a person checking nonclinical appointment logistics, begin with a calendar card and preparation instructions and do not disclose the intended interpretation. Withhold the source original and ask the reader, in their own language, to explain the message as if describing it to someone who has not seen the screen. The working question is whether the person can distinguish “arrival buffer and scheduled start are distinct” from the tempting claim that an early-arrival instruction is a different appointment time. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Listen for the causal story: what triggers the action, what stays true afterward and which actor controls the change. The expected observation is that the person identifies the correct appointment time and required document. If instead the answer implies that an early-arrival instruction is a different appointment time, log the exact words and where the reader looked. The potential harm is a missed appointment or inappropriate assumptions about professional advice. Do not treat a fluent reformulation as proof of meaning; a logistical instruction should not turn into domain advice. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to show both times and avoid adding advice not present in the source. A clean paraphrase should preserve the decision rule rather than repeat isolated keywords. A polished but incorrect summary is more informative than a low-confidence guess because it shows the exact rule the reader has constructed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 038: Appointment preparation message × Task simulation

Reader-facing test. For a person checking nonclinical appointment logistics, begin with a calendar card and preparation instructions and do not disclose the intended interpretation. Give a fictional scenario with a choice between acting now, waiting for confirmation and searching for another condition; allow the reader to use only the visible translated material. The working question is whether the person can distinguish “arrival buffer and scheduled start are distinct” from the tempting claim that an early-arrival instruction is a different appointment time. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Capture the chosen action, the text used to justify it and whether a wrong branch seems attractive. The expected observation is that the person identifies the correct appointment time and required document. If instead the answer implies that an early-arrival instruction is a different appointment time, log the exact words and where the reader looked. The potential harm is a missed appointment or inappropriate assumptions about professional advice. Do not treat a fluent reformulation as proof of meaning; a logistical instruction should not turn into domain advice. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to show both times and avoid adding advice not present in the source. Do not score the final click alone. Someone may guess correctly from placement while holding the wrong belief. A complete result includes the reason, expected consequence and what additional information the person believes is needed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 039: Appointment preparation message × Think-aloud trace

Reader-facing test. For a person checking nonclinical appointment logistics, begin with a calendar card and preparation instructions and do not disclose the intended interpretation. Ask the reader to narrate thoughts while moving from the notice to the relevant action, but do not correct terminology or explain the organization’s policy. The working question is whether the person can distinguish “arrival buffer and scheduled start are distinct” from the tempting claim that an early-arrival instruction is a different appointment time. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Mark the moment a term is interpreted, the moment confidence changes and the point at which layout or icons redirect attention. The expected observation is that the person identifies the correct appointment time and required document. If instead the answer implies that an early-arrival instruction is a different appointment time, log the exact words and where the reader looked. The potential harm is a missed appointment or inappropriate assumptions about professional advice. Do not treat a fluent reformulation as proof of meaning; a logistical instruction should not turn into domain advice. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to show both times and avoid adding advice not present in the source. This method can identify an interface cue rather than a translation error. The very act of thinking aloud may change behavior, so report it as qualitative diagnostic evidence rather than a naturalistic time-to-completion measurement. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 040: Appointment preparation message × Contrastive boundary pair

Reader-facing test. For a person checking nonclinical appointment logistics, begin with a calendar card and preparation instructions and do not disclose the intended interpretation. Present two nearly identical fictional cases that differ only in the decisive condition; ask whether the outcome changes and why. The working question is whether the person can distinguish “arrival buffer and scheduled start are distinct” from the tempting claim that an early-arrival instruction is a different appointment time. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Look for the reader’s treatment of a status, exception or timing boundary that the source intended to preserve. The expected observation is that the person identifies the correct appointment time and required document. If instead the answer implies that an early-arrival instruction is a different appointment time, log the exact words and where the reader looked. The potential harm is a missed appointment or inappropriate assumptions about professional advice. Do not treat a fluent reformulation as proof of meaning; a logistical instruction should not turn into domain advice. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to show both times and avoid adding advice not present in the source. A reader who gives the same answer to both cases may have lost the qualifier. However, if the source rule itself is ambiguous, pause the test and clarify it with the content owner. Do not force respondents to guess at a rule the author has not specified. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 041: Appointment preparation message × Actor and responsibility map

Reader-facing test. For a person checking nonclinical appointment logistics, begin with a calendar card and preparation instructions and do not disclose the intended interpretation. Ask the reader to identify who submits, who decides, who can revise the decision and who is affected; let them point to the translated words supporting each answer. The working question is whether the person can distinguish “arrival buffer and scheduled start are distinct” from the tempting claim that an early-arrival instruction is a different appointment time. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare the named actors with the interpretation contract and note any responsibility the reader invents. The expected observation is that the person identifies the correct appointment time and required document. If instead the answer implies that an early-arrival instruction is a different appointment time, log the exact words and where the reader looked. The potential harm is a missed appointment or inappropriate assumptions about professional advice. Do not treat a fluent reformulation as proof of meaning; a logistical instruction should not turn into domain advice. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to show both times and avoid adding advice not present in the source. Passive voice is not automatically wrong, and some languages omit subjects naturally. The test is whether the actual actor is recoverable. If the reader supplies an unintended actor, identify whether nearby labels, prior notifications or grammatical choices created that inference. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 042: Appointment preparation message × Counterexample challenge

Reader-facing test. For a person checking nonclinical appointment logistics, begin with a calendar card and preparation instructions and do not disclose the intended interpretation. Invite the reader to invent an example in which the notice would not apply, or to say what evidence would make the current instruction no longer valid. The working question is whether the person can distinguish “arrival buffer and scheduled start are distinct” from the tempting claim that an early-arrival instruction is a different appointment time. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Inspect whether the reader recognizes limits, exceptions, unknown information and provisional states. The expected observation is that the person identifies the correct appointment time and required document. If instead the answer implies that an early-arrival instruction is a different appointment time, log the exact words and where the reader looked. The potential harm is a missed appointment or inappropriate assumptions about professional advice. Do not treat a fluent reformulation as proof of meaning; a logistical instruction should not turn into domain advice. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to show both times and avoid adding advice not present in the source. An inability to imagine a counterexample may be a testing skill issue, so offer a neutral hypothetical rather than marking immediate failure. The key evidence is the reader’s understanding of the boundary after an ordinary scenario is supplied without revealing the answer. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 043: Appointment preparation message × Delayed teach-back

Reader-facing test. For a person checking nonclinical appointment logistics, begin with a calendar card and preparation instructions and do not disclose the intended interpretation. After an unrelated short filler activity, ask the reader to explain the next step to a fictional colleague who cannot view the translated message. The working question is whether the person can distinguish “arrival buffer and scheduled start are distinct” from the tempting claim that an early-arrival instruction is a different appointment time. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Note which conditions survive memory, which qualifiers disappear and whether confidence rises as detail is lost. The expected observation is that the person identifies the correct appointment time and required document. If instead the answer implies that an early-arrival instruction is a different appointment time, log the exact words and where the reader looked. The potential harm is a missed appointment or inappropriate assumptions about professional advice. Do not treat a fluent reformulation as proof of meaning; a logistical instruction should not turn into domain advice. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to show both times and avoid adding advice not present in the source. Teach-back tests retained practical meaning, not memorization of exact wording. Compare it with first-exposure paraphrase. If the reader initially understood but later drops a critical exception, consider information placement, a reminder, a concise summary or a clearer call to action. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 044: Appointment preparation message × Alternative wording comparison

Reader-facing test. For a person checking nonclinical appointment logistics, begin with a calendar card and preparation instructions and do not disclose the intended interpretation. Show two approved candidate renderings in a balanced order across participants and ask what each would lead them to do; avoid asking which sounds more elegant. The working question is whether the person can distinguish “arrival buffer and scheduled start are distinct” from the tempting claim that an early-arrival instruction is a different appointment time. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Record whether the alternatives produce different real-world predictions despite similar fluency judgments. The expected observation is that the person identifies the correct appointment time and required document. If instead the answer implies that an early-arrival instruction is a different appointment time, log the exact words and where the reader looked. The potential harm is a missed appointment or inappropriate assumptions about professional advice. Do not treat a fluent reformulation as proof of meaning; a logistical instruction should not turn into domain advice. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to show both times and avoid adding advice not present in the source. Do not optimize language by majority preference alone: a popular label can still induce a serious wrong decision. Preference is a supporting measure. Select wording that better preserves the rule, is natural for the intended audience and survives a subsequent task-based retest. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 045: Appointment preparation message × Release-readiness retest

Reader-facing test. For a person checking nonclinical appointment logistics, begin with a calendar card and preparation instructions and do not disclose the intended interpretation. Present the repaired artifact with the same surrounding interface but an independent fictional case; recruit someone not coached by the earlier test where feasible. The working question is whether the person can distinguish “arrival buffer and scheduled start are distinct” from the tempting claim that an early-arrival instruction is a different appointment time. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare unprompted decisions, explanations and remaining errors rather than merely tallying favorable reactions. The expected observation is that the person identifies the correct appointment time and required document. If instead the answer implies that an early-arrival instruction is a different appointment time, log the exact words and where the reader looked. The potential harm is a missed appointment or inappropriate assumptions about professional advice. Do not treat a fluent reformulation as proof of meaning; a logistical instruction should not turn into domain advice. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to show both times and avoid adding advice not present in the source. A corrected phrase is not validated by the editor’s confidence. Preserve before-and-after versions, source approval, exact prompts, reader evidence and unresolved risks. Report that the tested failure diminished under these conditions, not that the translation is universally impossible to misunderstand. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Domain studio 6: Workplace equipment lock notice

A an authorized worker reading a machine status panel encounters a maintenance banner above a disabled control. The intended rule is: The equipment is unavailable while maintenance is recorded; only the designated clearance process changes that state. The risky alternative is a green-looking screen means the equipment is cleared for use. This matters because unsafe reliance on an unapproved state. Here the semantic hinge is that visual appearance and authorization state can diverge. These are invented teaching fixtures, so a real publisher must replace them with an approved source contract and its own audience research.

Diagnostic case 046: Workplace equipment lock notice × Blind paraphrase

Reader-facing test. For an authorized worker reading a machine status panel, begin with a maintenance banner above a disabled control and do not disclose the intended interpretation. Withhold the source original and ask the reader, in their own language, to explain the message as if describing it to someone who has not seen the screen. The working question is whether the person can distinguish “visual appearance and authorization state can diverge” from the tempting claim that a green-looking screen means the equipment is cleared for use. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Listen for the causal story: what triggers the action, what stays true afterward and which actor controls the change. The expected observation is that the person does not infer permission from the appearance of a button. If instead the answer implies that a green-looking screen means the equipment is cleared for use, log the exact words and where the reader looked. The potential harm is unsafe reliance on an unapproved state. Do not treat a fluent reformulation as proof of meaning; state names carry consequences beyond their visual form. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to make the authorization status and responsible role unmistakable. A clean paraphrase should preserve the decision rule rather than repeat isolated keywords. A polished but incorrect summary is more informative than a low-confidence guess because it shows the exact rule the reader has constructed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 047: Workplace equipment lock notice × Task simulation

Reader-facing test. For an authorized worker reading a machine status panel, begin with a maintenance banner above a disabled control and do not disclose the intended interpretation. Give a fictional scenario with a choice between acting now, waiting for confirmation and searching for another condition; allow the reader to use only the visible translated material. The working question is whether the person can distinguish “visual appearance and authorization state can diverge” from the tempting claim that a green-looking screen means the equipment is cleared for use. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Capture the chosen action, the text used to justify it and whether a wrong branch seems attractive. The expected observation is that the person does not infer permission from the appearance of a button. If instead the answer implies that a green-looking screen means the equipment is cleared for use, log the exact words and where the reader looked. The potential harm is unsafe reliance on an unapproved state. Do not treat a fluent reformulation as proof of meaning; state names carry consequences beyond their visual form. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to make the authorization status and responsible role unmistakable. Do not score the final click alone. Someone may guess correctly from placement while holding the wrong belief. A complete result includes the reason, expected consequence and what additional information the person believes is needed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 048: Workplace equipment lock notice × Think-aloud trace

Reader-facing test. For an authorized worker reading a machine status panel, begin with a maintenance banner above a disabled control and do not disclose the intended interpretation. Ask the reader to narrate thoughts while moving from the notice to the relevant action, but do not correct terminology or explain the organization’s policy. The working question is whether the person can distinguish “visual appearance and authorization state can diverge” from the tempting claim that a green-looking screen means the equipment is cleared for use. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Mark the moment a term is interpreted, the moment confidence changes and the point at which layout or icons redirect attention. The expected observation is that the person does not infer permission from the appearance of a button. If instead the answer implies that a green-looking screen means the equipment is cleared for use, log the exact words and where the reader looked. The potential harm is unsafe reliance on an unapproved state. Do not treat a fluent reformulation as proof of meaning; state names carry consequences beyond their visual form. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to make the authorization status and responsible role unmistakable. This method can identify an interface cue rather than a translation error. The very act of thinking aloud may change behavior, so report it as qualitative diagnostic evidence rather than a naturalistic time-to-completion measurement. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 049: Workplace equipment lock notice × Contrastive boundary pair

Reader-facing test. For an authorized worker reading a machine status panel, begin with a maintenance banner above a disabled control and do not disclose the intended interpretation. Present two nearly identical fictional cases that differ only in the decisive condition; ask whether the outcome changes and why. The working question is whether the person can distinguish “visual appearance and authorization state can diverge” from the tempting claim that a green-looking screen means the equipment is cleared for use. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Look for the reader’s treatment of a status, exception or timing boundary that the source intended to preserve. The expected observation is that the person does not infer permission from the appearance of a button. If instead the answer implies that a green-looking screen means the equipment is cleared for use, log the exact words and where the reader looked. The potential harm is unsafe reliance on an unapproved state. Do not treat a fluent reformulation as proof of meaning; state names carry consequences beyond their visual form. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to make the authorization status and responsible role unmistakable. A reader who gives the same answer to both cases may have lost the qualifier. However, if the source rule itself is ambiguous, pause the test and clarify it with the content owner. Do not force respondents to guess at a rule the author has not specified. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 050: Workplace equipment lock notice × Actor and responsibility map

Reader-facing test. For an authorized worker reading a machine status panel, begin with a maintenance banner above a disabled control and do not disclose the intended interpretation. Ask the reader to identify who submits, who decides, who can revise the decision and who is affected; let them point to the translated words supporting each answer. The working question is whether the person can distinguish “visual appearance and authorization state can diverge” from the tempting claim that a green-looking screen means the equipment is cleared for use. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare the named actors with the interpretation contract and note any responsibility the reader invents. The expected observation is that the person does not infer permission from the appearance of a button. If instead the answer implies that a green-looking screen means the equipment is cleared for use, log the exact words and where the reader looked. The potential harm is unsafe reliance on an unapproved state. Do not treat a fluent reformulation as proof of meaning; state names carry consequences beyond their visual form. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to make the authorization status and responsible role unmistakable. Passive voice is not automatically wrong, and some languages omit subjects naturally. The test is whether the actual actor is recoverable. If the reader supplies an unintended actor, identify whether nearby labels, prior notifications or grammatical choices created that inference. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 051: Workplace equipment lock notice × Counterexample challenge

Reader-facing test. For an authorized worker reading a machine status panel, begin with a maintenance banner above a disabled control and do not disclose the intended interpretation. Invite the reader to invent an example in which the notice would not apply, or to say what evidence would make the current instruction no longer valid. The working question is whether the person can distinguish “visual appearance and authorization state can diverge” from the tempting claim that a green-looking screen means the equipment is cleared for use. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Inspect whether the reader recognizes limits, exceptions, unknown information and provisional states. The expected observation is that the person does not infer permission from the appearance of a button. If instead the answer implies that a green-looking screen means the equipment is cleared for use, log the exact words and where the reader looked. The potential harm is unsafe reliance on an unapproved state. Do not treat a fluent reformulation as proof of meaning; state names carry consequences beyond their visual form. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to make the authorization status and responsible role unmistakable. An inability to imagine a counterexample may be a testing skill issue, so offer a neutral hypothetical rather than marking immediate failure. The key evidence is the reader’s understanding of the boundary after an ordinary scenario is supplied without revealing the answer. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 052: Workplace equipment lock notice × Delayed teach-back

Reader-facing test. For an authorized worker reading a machine status panel, begin with a maintenance banner above a disabled control and do not disclose the intended interpretation. After an unrelated short filler activity, ask the reader to explain the next step to a fictional colleague who cannot view the translated message. The working question is whether the person can distinguish “visual appearance and authorization state can diverge” from the tempting claim that a green-looking screen means the equipment is cleared for use. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Note which conditions survive memory, which qualifiers disappear and whether confidence rises as detail is lost. The expected observation is that the person does not infer permission from the appearance of a button. If instead the answer implies that a green-looking screen means the equipment is cleared for use, log the exact words and where the reader looked. The potential harm is unsafe reliance on an unapproved state. Do not treat a fluent reformulation as proof of meaning; state names carry consequences beyond their visual form. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to make the authorization status and responsible role unmistakable. Teach-back tests retained practical meaning, not memorization of exact wording. Compare it with first-exposure paraphrase. If the reader initially understood but later drops a critical exception, consider information placement, a reminder, a concise summary or a clearer call to action. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 053: Workplace equipment lock notice × Alternative wording comparison

Reader-facing test. For an authorized worker reading a machine status panel, begin with a maintenance banner above a disabled control and do not disclose the intended interpretation. Show two approved candidate renderings in a balanced order across participants and ask what each would lead them to do; avoid asking which sounds more elegant. The working question is whether the person can distinguish “visual appearance and authorization state can diverge” from the tempting claim that a green-looking screen means the equipment is cleared for use. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Record whether the alternatives produce different real-world predictions despite similar fluency judgments. The expected observation is that the person does not infer permission from the appearance of a button. If instead the answer implies that a green-looking screen means the equipment is cleared for use, log the exact words and where the reader looked. The potential harm is unsafe reliance on an unapproved state. Do not treat a fluent reformulation as proof of meaning; state names carry consequences beyond their visual form. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to make the authorization status and responsible role unmistakable. Do not optimize language by majority preference alone: a popular label can still induce a serious wrong decision. Preference is a supporting measure. Select wording that better preserves the rule, is natural for the intended audience and survives a subsequent task-based retest. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 054: Workplace equipment lock notice × Release-readiness retest

Reader-facing test. For an authorized worker reading a machine status panel, begin with a maintenance banner above a disabled control and do not disclose the intended interpretation. Present the repaired artifact with the same surrounding interface but an independent fictional case; recruit someone not coached by the earlier test where feasible. The working question is whether the person can distinguish “visual appearance and authorization state can diverge” from the tempting claim that a green-looking screen means the equipment is cleared for use. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare unprompted decisions, explanations and remaining errors rather than merely tallying favorable reactions. The expected observation is that the person does not infer permission from the appearance of a button. If instead the answer implies that a green-looking screen means the equipment is cleared for use, log the exact words and where the reader looked. The potential harm is unsafe reliance on an unapproved state. Do not treat a fluent reformulation as proof of meaning; state names carry consequences beyond their visual form. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to make the authorization status and responsible role unmistakable. A corrected phrase is not validated by the editor’s confidence. Preserve before-and-after versions, source approval, exact prompts, reader evidence and unresolved risks. Report that the tested failure diminished under these conditions, not that the translation is universally impossible to misunderstand. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Domain studio 7: Emergency assembly direction

A a visitor reading an incident-information card encounters a map panel with a closure notification. The intended rule is: The named meeting point applies only when the posted route is open; a closed route requires following current official directions. The risky alternative is the named meeting point must always be reached by the original path. This matters because people follow an unavailable path. Here the semantic hinge is that destination and route availability have separate conditions. These are invented teaching fixtures, so a real publisher must replace them with an approved source contract and its own audience research.

Diagnostic case 055: Emergency assembly direction × Blind paraphrase

Reader-facing test. For a visitor reading an incident-information card, begin with a map panel with a closure notification and do not disclose the intended interpretation. Withhold the source original and ask the reader, in their own language, to explain the message as if describing it to someone who has not seen the screen. The working question is whether the person can distinguish “destination and route availability have separate conditions” from the tempting claim that the named meeting point must always be reached by the original path. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Listen for the causal story: what triggers the action, what stays true afterward and which actor controls the change. The expected observation is that the person explains how a closure changes route selection without inventing one. If instead the answer implies that the named meeting point must always be reached by the original path, log the exact words and where the reader looked. The potential harm is people follow an unavailable path. Do not treat a fluent reformulation as proof of meaning; conditional public guidance is not a universal route guarantee. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to pair the route condition with the destination instruction. A clean paraphrase should preserve the decision rule rather than repeat isolated keywords. A polished but incorrect summary is more informative than a low-confidence guess because it shows the exact rule the reader has constructed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 056: Emergency assembly direction × Task simulation

Reader-facing test. For a visitor reading an incident-information card, begin with a map panel with a closure notification and do not disclose the intended interpretation. Give a fictional scenario with a choice between acting now, waiting for confirmation and searching for another condition; allow the reader to use only the visible translated material. The working question is whether the person can distinguish “destination and route availability have separate conditions” from the tempting claim that the named meeting point must always be reached by the original path. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Capture the chosen action, the text used to justify it and whether a wrong branch seems attractive. The expected observation is that the person explains how a closure changes route selection without inventing one. If instead the answer implies that the named meeting point must always be reached by the original path, log the exact words and where the reader looked. The potential harm is people follow an unavailable path. Do not treat a fluent reformulation as proof of meaning; conditional public guidance is not a universal route guarantee. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to pair the route condition with the destination instruction. Do not score the final click alone. Someone may guess correctly from placement while holding the wrong belief. A complete result includes the reason, expected consequence and what additional information the person believes is needed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 057: Emergency assembly direction × Think-aloud trace

Reader-facing test. For a visitor reading an incident-information card, begin with a map panel with a closure notification and do not disclose the intended interpretation. Ask the reader to narrate thoughts while moving from the notice to the relevant action, but do not correct terminology or explain the organization’s policy. The working question is whether the person can distinguish “destination and route availability have separate conditions” from the tempting claim that the named meeting point must always be reached by the original path. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Mark the moment a term is interpreted, the moment confidence changes and the point at which layout or icons redirect attention. The expected observation is that the person explains how a closure changes route selection without inventing one. If instead the answer implies that the named meeting point must always be reached by the original path, log the exact words and where the reader looked. The potential harm is people follow an unavailable path. Do not treat a fluent reformulation as proof of meaning; conditional public guidance is not a universal route guarantee. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to pair the route condition with the destination instruction. This method can identify an interface cue rather than a translation error. The very act of thinking aloud may change behavior, so report it as qualitative diagnostic evidence rather than a naturalistic time-to-completion measurement. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 058: Emergency assembly direction × Contrastive boundary pair

Reader-facing test. For a visitor reading an incident-information card, begin with a map panel with a closure notification and do not disclose the intended interpretation. Present two nearly identical fictional cases that differ only in the decisive condition; ask whether the outcome changes and why. The working question is whether the person can distinguish “destination and route availability have separate conditions” from the tempting claim that the named meeting point must always be reached by the original path. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Look for the reader’s treatment of a status, exception or timing boundary that the source intended to preserve. The expected observation is that the person explains how a closure changes route selection without inventing one. If instead the answer implies that the named meeting point must always be reached by the original path, log the exact words and where the reader looked. The potential harm is people follow an unavailable path. Do not treat a fluent reformulation as proof of meaning; conditional public guidance is not a universal route guarantee. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to pair the route condition with the destination instruction. A reader who gives the same answer to both cases may have lost the qualifier. However, if the source rule itself is ambiguous, pause the test and clarify it with the content owner. Do not force respondents to guess at a rule the author has not specified. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 059: Emergency assembly direction × Actor and responsibility map

Reader-facing test. For a visitor reading an incident-information card, begin with a map panel with a closure notification and do not disclose the intended interpretation. Ask the reader to identify who submits, who decides, who can revise the decision and who is affected; let them point to the translated words supporting each answer. The working question is whether the person can distinguish “destination and route availability have separate conditions” from the tempting claim that the named meeting point must always be reached by the original path. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare the named actors with the interpretation contract and note any responsibility the reader invents. The expected observation is that the person explains how a closure changes route selection without inventing one. If instead the answer implies that the named meeting point must always be reached by the original path, log the exact words and where the reader looked. The potential harm is people follow an unavailable path. Do not treat a fluent reformulation as proof of meaning; conditional public guidance is not a universal route guarantee. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to pair the route condition with the destination instruction. Passive voice is not automatically wrong, and some languages omit subjects naturally. The test is whether the actual actor is recoverable. If the reader supplies an unintended actor, identify whether nearby labels, prior notifications or grammatical choices created that inference. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 060: Emergency assembly direction × Counterexample challenge

Reader-facing test. For a visitor reading an incident-information card, begin with a map panel with a closure notification and do not disclose the intended interpretation. Invite the reader to invent an example in which the notice would not apply, or to say what evidence would make the current instruction no longer valid. The working question is whether the person can distinguish “destination and route availability have separate conditions” from the tempting claim that the named meeting point must always be reached by the original path. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Inspect whether the reader recognizes limits, exceptions, unknown information and provisional states. The expected observation is that the person explains how a closure changes route selection without inventing one. If instead the answer implies that the named meeting point must always be reached by the original path, log the exact words and where the reader looked. The potential harm is people follow an unavailable path. Do not treat a fluent reformulation as proof of meaning; conditional public guidance is not a universal route guarantee. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to pair the route condition with the destination instruction. An inability to imagine a counterexample may be a testing skill issue, so offer a neutral hypothetical rather than marking immediate failure. The key evidence is the reader’s understanding of the boundary after an ordinary scenario is supplied without revealing the answer. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 061: Emergency assembly direction × Delayed teach-back

Reader-facing test. For a visitor reading an incident-information card, begin with a map panel with a closure notification and do not disclose the intended interpretation. After an unrelated short filler activity, ask the reader to explain the next step to a fictional colleague who cannot view the translated message. The working question is whether the person can distinguish “destination and route availability have separate conditions” from the tempting claim that the named meeting point must always be reached by the original path. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Note which conditions survive memory, which qualifiers disappear and whether confidence rises as detail is lost. The expected observation is that the person explains how a closure changes route selection without inventing one. If instead the answer implies that the named meeting point must always be reached by the original path, log the exact words and where the reader looked. The potential harm is people follow an unavailable path. Do not treat a fluent reformulation as proof of meaning; conditional public guidance is not a universal route guarantee. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to pair the route condition with the destination instruction. Teach-back tests retained practical meaning, not memorization of exact wording. Compare it with first-exposure paraphrase. If the reader initially understood but later drops a critical exception, consider information placement, a reminder, a concise summary or a clearer call to action. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 062: Emergency assembly direction × Alternative wording comparison

Reader-facing test. For a visitor reading an incident-information card, begin with a map panel with a closure notification and do not disclose the intended interpretation. Show two approved candidate renderings in a balanced order across participants and ask what each would lead them to do; avoid asking which sounds more elegant. The working question is whether the person can distinguish “destination and route availability have separate conditions” from the tempting claim that the named meeting point must always be reached by the original path. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Record whether the alternatives produce different real-world predictions despite similar fluency judgments. The expected observation is that the person explains how a closure changes route selection without inventing one. If instead the answer implies that the named meeting point must always be reached by the original path, log the exact words and where the reader looked. The potential harm is people follow an unavailable path. Do not treat a fluent reformulation as proof of meaning; conditional public guidance is not a universal route guarantee. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to pair the route condition with the destination instruction. Do not optimize language by majority preference alone: a popular label can still induce a serious wrong decision. Preference is a supporting measure. Select wording that better preserves the rule, is natural for the intended audience and survives a subsequent task-based retest. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 063: Emergency assembly direction × Release-readiness retest

Reader-facing test. For a visitor reading an incident-information card, begin with a map panel with a closure notification and do not disclose the intended interpretation. Present the repaired artifact with the same surrounding interface but an independent fictional case; recruit someone not coached by the earlier test where feasible. The working question is whether the person can distinguish “destination and route availability have separate conditions” from the tempting claim that the named meeting point must always be reached by the original path. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare unprompted decisions, explanations and remaining errors rather than merely tallying favorable reactions. The expected observation is that the person explains how a closure changes route selection without inventing one. If instead the answer implies that the named meeting point must always be reached by the original path, log the exact words and where the reader looked. The potential harm is people follow an unavailable path. Do not treat a fluent reformulation as proof of meaning; conditional public guidance is not a universal route guarantee. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to pair the route condition with the destination instruction. A corrected phrase is not validated by the editor’s confidence. Preserve before-and-after versions, source approval, exact prompts, reader evidence and unresolved risks. Report that the tested failure diminished under these conditions, not that the translation is universally impossible to misunderstand. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Domain studio 8: Travel service disruption

A a passenger reviewing itinerary updates encounters a new itinerary card above an unchanged booking reference. The intended rule is: The scheduled departure has changed, but the reservation is not confirmed on the replacement service until the operator issues confirmation. The risky alternative is a listed alternative is an automatic confirmed booking. This matters because missed travel or unfounded reliance on an alternative. Here the semantic hinge is that options and commitments have different states. These are invented teaching fixtures, so a real publisher must replace them with an approved source contract and its own audience research.

Diagnostic case 064: Travel service disruption × Blind paraphrase

Reader-facing test. For a passenger reviewing itinerary updates, begin with a new itinerary card above an unchanged booking reference and do not disclose the intended interpretation. Withhold the source original and ask the reader, in their own language, to explain the message as if describing it to someone who has not seen the screen. The working question is whether the person can distinguish “options and commitments have different states” from the tempting claim that a listed alternative is an automatic confirmed booking. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Listen for the causal story: what triggers the action, what stays true afterward and which actor controls the change. The expected observation is that the person distinguishes information about a service from a confirmed seat. If instead the answer implies that a listed alternative is an automatic confirmed booking, log the exact words and where the reader looked. The potential harm is missed travel or unfounded reliance on an alternative. Do not treat a fluent reformulation as proof of meaning; a suggestion does not become an entitlement. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label offered alternatives as pending until confirmed. A clean paraphrase should preserve the decision rule rather than repeat isolated keywords. A polished but incorrect summary is more informative than a low-confidence guess because it shows the exact rule the reader has constructed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 065: Travel service disruption × Task simulation

Reader-facing test. For a passenger reviewing itinerary updates, begin with a new itinerary card above an unchanged booking reference and do not disclose the intended interpretation. Give a fictional scenario with a choice between acting now, waiting for confirmation and searching for another condition; allow the reader to use only the visible translated material. The working question is whether the person can distinguish “options and commitments have different states” from the tempting claim that a listed alternative is an automatic confirmed booking. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Capture the chosen action, the text used to justify it and whether a wrong branch seems attractive. The expected observation is that the person distinguishes information about a service from a confirmed seat. If instead the answer implies that a listed alternative is an automatic confirmed booking, log the exact words and where the reader looked. The potential harm is missed travel or unfounded reliance on an alternative. Do not treat a fluent reformulation as proof of meaning; a suggestion does not become an entitlement. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label offered alternatives as pending until confirmed. Do not score the final click alone. Someone may guess correctly from placement while holding the wrong belief. A complete result includes the reason, expected consequence and what additional information the person believes is needed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 066: Travel service disruption × Think-aloud trace

Reader-facing test. For a passenger reviewing itinerary updates, begin with a new itinerary card above an unchanged booking reference and do not disclose the intended interpretation. Ask the reader to narrate thoughts while moving from the notice to the relevant action, but do not correct terminology or explain the organization’s policy. The working question is whether the person can distinguish “options and commitments have different states” from the tempting claim that a listed alternative is an automatic confirmed booking. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Mark the moment a term is interpreted, the moment confidence changes and the point at which layout or icons redirect attention. The expected observation is that the person distinguishes information about a service from a confirmed seat. If instead the answer implies that a listed alternative is an automatic confirmed booking, log the exact words and where the reader looked. The potential harm is missed travel or unfounded reliance on an alternative. Do not treat a fluent reformulation as proof of meaning; a suggestion does not become an entitlement. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label offered alternatives as pending until confirmed. This method can identify an interface cue rather than a translation error. The very act of thinking aloud may change behavior, so report it as qualitative diagnostic evidence rather than a naturalistic time-to-completion measurement. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 067: Travel service disruption × Contrastive boundary pair

Reader-facing test. For a passenger reviewing itinerary updates, begin with a new itinerary card above an unchanged booking reference and do not disclose the intended interpretation. Present two nearly identical fictional cases that differ only in the decisive condition; ask whether the outcome changes and why. The working question is whether the person can distinguish “options and commitments have different states” from the tempting claim that a listed alternative is an automatic confirmed booking. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Look for the reader’s treatment of a status, exception or timing boundary that the source intended to preserve. The expected observation is that the person distinguishes information about a service from a confirmed seat. If instead the answer implies that a listed alternative is an automatic confirmed booking, log the exact words and where the reader looked. The potential harm is missed travel or unfounded reliance on an alternative. Do not treat a fluent reformulation as proof of meaning; a suggestion does not become an entitlement. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label offered alternatives as pending until confirmed. A reader who gives the same answer to both cases may have lost the qualifier. However, if the source rule itself is ambiguous, pause the test and clarify it with the content owner. Do not force respondents to guess at a rule the author has not specified. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 068: Travel service disruption × Actor and responsibility map

Reader-facing test. For a passenger reviewing itinerary updates, begin with a new itinerary card above an unchanged booking reference and do not disclose the intended interpretation. Ask the reader to identify who submits, who decides, who can revise the decision and who is affected; let them point to the translated words supporting each answer. The working question is whether the person can distinguish “options and commitments have different states” from the tempting claim that a listed alternative is an automatic confirmed booking. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare the named actors with the interpretation contract and note any responsibility the reader invents. The expected observation is that the person distinguishes information about a service from a confirmed seat. If instead the answer implies that a listed alternative is an automatic confirmed booking, log the exact words and where the reader looked. The potential harm is missed travel or unfounded reliance on an alternative. Do not treat a fluent reformulation as proof of meaning; a suggestion does not become an entitlement. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label offered alternatives as pending until confirmed. Passive voice is not automatically wrong, and some languages omit subjects naturally. The test is whether the actual actor is recoverable. If the reader supplies an unintended actor, identify whether nearby labels, prior notifications or grammatical choices created that inference. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 069: Travel service disruption × Counterexample challenge

Reader-facing test. For a passenger reviewing itinerary updates, begin with a new itinerary card above an unchanged booking reference and do not disclose the intended interpretation. Invite the reader to invent an example in which the notice would not apply, or to say what evidence would make the current instruction no longer valid. The working question is whether the person can distinguish “options and commitments have different states” from the tempting claim that a listed alternative is an automatic confirmed booking. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Inspect whether the reader recognizes limits, exceptions, unknown information and provisional states. The expected observation is that the person distinguishes information about a service from a confirmed seat. If instead the answer implies that a listed alternative is an automatic confirmed booking, log the exact words and where the reader looked. The potential harm is missed travel or unfounded reliance on an alternative. Do not treat a fluent reformulation as proof of meaning; a suggestion does not become an entitlement. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label offered alternatives as pending until confirmed. An inability to imagine a counterexample may be a testing skill issue, so offer a neutral hypothetical rather than marking immediate failure. The key evidence is the reader’s understanding of the boundary after an ordinary scenario is supplied without revealing the answer. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 070: Travel service disruption × Delayed teach-back

Reader-facing test. For a passenger reviewing itinerary updates, begin with a new itinerary card above an unchanged booking reference and do not disclose the intended interpretation. After an unrelated short filler activity, ask the reader to explain the next step to a fictional colleague who cannot view the translated message. The working question is whether the person can distinguish “options and commitments have different states” from the tempting claim that a listed alternative is an automatic confirmed booking. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Note which conditions survive memory, which qualifiers disappear and whether confidence rises as detail is lost. The expected observation is that the person distinguishes information about a service from a confirmed seat. If instead the answer implies that a listed alternative is an automatic confirmed booking, log the exact words and where the reader looked. The potential harm is missed travel or unfounded reliance on an alternative. Do not treat a fluent reformulation as proof of meaning; a suggestion does not become an entitlement. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label offered alternatives as pending until confirmed. Teach-back tests retained practical meaning, not memorization of exact wording. Compare it with first-exposure paraphrase. If the reader initially understood but later drops a critical exception, consider information placement, a reminder, a concise summary or a clearer call to action. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 071: Travel service disruption × Alternative wording comparison

Reader-facing test. For a passenger reviewing itinerary updates, begin with a new itinerary card above an unchanged booking reference and do not disclose the intended interpretation. Show two approved candidate renderings in a balanced order across participants and ask what each would lead them to do; avoid asking which sounds more elegant. The working question is whether the person can distinguish “options and commitments have different states” from the tempting claim that a listed alternative is an automatic confirmed booking. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Record whether the alternatives produce different real-world predictions despite similar fluency judgments. The expected observation is that the person distinguishes information about a service from a confirmed seat. If instead the answer implies that a listed alternative is an automatic confirmed booking, log the exact words and where the reader looked. The potential harm is missed travel or unfounded reliance on an alternative. Do not treat a fluent reformulation as proof of meaning; a suggestion does not become an entitlement. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label offered alternatives as pending until confirmed. Do not optimize language by majority preference alone: a popular label can still induce a serious wrong decision. Preference is a supporting measure. Select wording that better preserves the rule, is natural for the intended audience and survives a subsequent task-based retest. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 072: Travel service disruption × Release-readiness retest

Reader-facing test. For a passenger reviewing itinerary updates, begin with a new itinerary card above an unchanged booking reference and do not disclose the intended interpretation. Present the repaired artifact with the same surrounding interface but an independent fictional case; recruit someone not coached by the earlier test where feasible. The working question is whether the person can distinguish “options and commitments have different states” from the tempting claim that a listed alternative is an automatic confirmed booking. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare unprompted decisions, explanations and remaining errors rather than merely tallying favorable reactions. The expected observation is that the person distinguishes information about a service from a confirmed seat. If instead the answer implies that a listed alternative is an automatic confirmed booking, log the exact words and where the reader looked. The potential harm is missed travel or unfounded reliance on an alternative. Do not treat a fluent reformulation as proof of meaning; a suggestion does not become an entitlement. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label offered alternatives as pending until confirmed. A corrected phrase is not validated by the editor’s confidence. Preserve before-and-after versions, source approval, exact prompts, reader evidence and unresolved risks. Report that the tested failure diminished under these conditions, not that the translation is universally impossible to misunderstand. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Domain studio 9: Library borrowing deadline

A a borrower looking at a due-date reminder encounters a renewal button and current due date. The intended rule is: A borrowed item is due on the shown day; a renewal request is not effective unless approved by the system. The risky alternative is clicking request renewal always extends the loan. This matters because late return based on an unapproved extension. Here the semantic hinge is that a request action does not equal a successful state transition. These are invented teaching fixtures, so a real publisher must replace them with an approved source contract and its own audience research.

Diagnostic case 073: Library borrowing deadline × Blind paraphrase

Reader-facing test. For a borrower looking at a due-date reminder, begin with a renewal button and current due date and do not disclose the intended interpretation. Withhold the source original and ask the reader, in their own language, to explain the message as if describing it to someone who has not seen the screen. The working question is whether the person can distinguish “a request action does not equal a successful state transition” from the tempting claim that clicking request renewal always extends the loan. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Listen for the causal story: what triggers the action, what stays true afterward and which actor controls the change. The expected observation is that the person states which date controls while the renewal is pending. If instead the answer implies that clicking request renewal always extends the loan, log the exact words and where the reader looked. The potential harm is late return based on an unapproved extension. Do not treat a fluent reformulation as proof of meaning; request, confirmation and effective change are three separate events. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to display the current deadline until confirmation arrives. A clean paraphrase should preserve the decision rule rather than repeat isolated keywords. A polished but incorrect summary is more informative than a low-confidence guess because it shows the exact rule the reader has constructed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 074: Library borrowing deadline × Task simulation

Reader-facing test. For a borrower looking at a due-date reminder, begin with a renewal button and current due date and do not disclose the intended interpretation. Give a fictional scenario with a choice between acting now, waiting for confirmation and searching for another condition; allow the reader to use only the visible translated material. The working question is whether the person can distinguish “a request action does not equal a successful state transition” from the tempting claim that clicking request renewal always extends the loan. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Capture the chosen action, the text used to justify it and whether a wrong branch seems attractive. The expected observation is that the person states which date controls while the renewal is pending. If instead the answer implies that clicking request renewal always extends the loan, log the exact words and where the reader looked. The potential harm is late return based on an unapproved extension. Do not treat a fluent reformulation as proof of meaning; request, confirmation and effective change are three separate events. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to display the current deadline until confirmation arrives. Do not score the final click alone. Someone may guess correctly from placement while holding the wrong belief. A complete result includes the reason, expected consequence and what additional information the person believes is needed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 075: Library borrowing deadline × Think-aloud trace

Reader-facing test. For a borrower looking at a due-date reminder, begin with a renewal button and current due date and do not disclose the intended interpretation. Ask the reader to narrate thoughts while moving from the notice to the relevant action, but do not correct terminology or explain the organization’s policy. The working question is whether the person can distinguish “a request action does not equal a successful state transition” from the tempting claim that clicking request renewal always extends the loan. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Mark the moment a term is interpreted, the moment confidence changes and the point at which layout or icons redirect attention. The expected observation is that the person states which date controls while the renewal is pending. If instead the answer implies that clicking request renewal always extends the loan, log the exact words and where the reader looked. The potential harm is late return based on an unapproved extension. Do not treat a fluent reformulation as proof of meaning; request, confirmation and effective change are three separate events. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to display the current deadline until confirmation arrives. This method can identify an interface cue rather than a translation error. The very act of thinking aloud may change behavior, so report it as qualitative diagnostic evidence rather than a naturalistic time-to-completion measurement. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 076: Library borrowing deadline × Contrastive boundary pair

Reader-facing test. For a borrower looking at a due-date reminder, begin with a renewal button and current due date and do not disclose the intended interpretation. Present two nearly identical fictional cases that differ only in the decisive condition; ask whether the outcome changes and why. The working question is whether the person can distinguish “a request action does not equal a successful state transition” from the tempting claim that clicking request renewal always extends the loan. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Look for the reader’s treatment of a status, exception or timing boundary that the source intended to preserve. The expected observation is that the person states which date controls while the renewal is pending. If instead the answer implies that clicking request renewal always extends the loan, log the exact words and where the reader looked. The potential harm is late return based on an unapproved extension. Do not treat a fluent reformulation as proof of meaning; request, confirmation and effective change are three separate events. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to display the current deadline until confirmation arrives. A reader who gives the same answer to both cases may have lost the qualifier. However, if the source rule itself is ambiguous, pause the test and clarify it with the content owner. Do not force respondents to guess at a rule the author has not specified. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 077: Library borrowing deadline × Actor and responsibility map

Reader-facing test. For a borrower looking at a due-date reminder, begin with a renewal button and current due date and do not disclose the intended interpretation. Ask the reader to identify who submits, who decides, who can revise the decision and who is affected; let them point to the translated words supporting each answer. The working question is whether the person can distinguish “a request action does not equal a successful state transition” from the tempting claim that clicking request renewal always extends the loan. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare the named actors with the interpretation contract and note any responsibility the reader invents. The expected observation is that the person states which date controls while the renewal is pending. If instead the answer implies that clicking request renewal always extends the loan, log the exact words and where the reader looked. The potential harm is late return based on an unapproved extension. Do not treat a fluent reformulation as proof of meaning; request, confirmation and effective change are three separate events. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to display the current deadline until confirmation arrives. Passive voice is not automatically wrong, and some languages omit subjects naturally. The test is whether the actual actor is recoverable. If the reader supplies an unintended actor, identify whether nearby labels, prior notifications or grammatical choices created that inference. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 078: Library borrowing deadline × Counterexample challenge

Reader-facing test. For a borrower looking at a due-date reminder, begin with a renewal button and current due date and do not disclose the intended interpretation. Invite the reader to invent an example in which the notice would not apply, or to say what evidence would make the current instruction no longer valid. The working question is whether the person can distinguish “a request action does not equal a successful state transition” from the tempting claim that clicking request renewal always extends the loan. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Inspect whether the reader recognizes limits, exceptions, unknown information and provisional states. The expected observation is that the person states which date controls while the renewal is pending. If instead the answer implies that clicking request renewal always extends the loan, log the exact words and where the reader looked. The potential harm is late return based on an unapproved extension. Do not treat a fluent reformulation as proof of meaning; request, confirmation and effective change are three separate events. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to display the current deadline until confirmation arrives. An inability to imagine a counterexample may be a testing skill issue, so offer a neutral hypothetical rather than marking immediate failure. The key evidence is the reader’s understanding of the boundary after an ordinary scenario is supplied without revealing the answer. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 079: Library borrowing deadline × Delayed teach-back

Reader-facing test. For a borrower looking at a due-date reminder, begin with a renewal button and current due date and do not disclose the intended interpretation. After an unrelated short filler activity, ask the reader to explain the next step to a fictional colleague who cannot view the translated message. The working question is whether the person can distinguish “a request action does not equal a successful state transition” from the tempting claim that clicking request renewal always extends the loan. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Note which conditions survive memory, which qualifiers disappear and whether confidence rises as detail is lost. The expected observation is that the person states which date controls while the renewal is pending. If instead the answer implies that clicking request renewal always extends the loan, log the exact words and where the reader looked. The potential harm is late return based on an unapproved extension. Do not treat a fluent reformulation as proof of meaning; request, confirmation and effective change are three separate events. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to display the current deadline until confirmation arrives. Teach-back tests retained practical meaning, not memorization of exact wording. Compare it with first-exposure paraphrase. If the reader initially understood but later drops a critical exception, consider information placement, a reminder, a concise summary or a clearer call to action. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 080: Library borrowing deadline × Alternative wording comparison

Reader-facing test. For a borrower looking at a due-date reminder, begin with a renewal button and current due date and do not disclose the intended interpretation. Show two approved candidate renderings in a balanced order across participants and ask what each would lead them to do; avoid asking which sounds more elegant. The working question is whether the person can distinguish “a request action does not equal a successful state transition” from the tempting claim that clicking request renewal always extends the loan. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Record whether the alternatives produce different real-world predictions despite similar fluency judgments. The expected observation is that the person states which date controls while the renewal is pending. If instead the answer implies that clicking request renewal always extends the loan, log the exact words and where the reader looked. The potential harm is late return based on an unapproved extension. Do not treat a fluent reformulation as proof of meaning; request, confirmation and effective change are three separate events. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to display the current deadline until confirmation arrives. Do not optimize language by majority preference alone: a popular label can still induce a serious wrong decision. Preference is a supporting measure. Select wording that better preserves the rule, is natural for the intended audience and survives a subsequent task-based retest. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 081: Library borrowing deadline × Release-readiness retest

Reader-facing test. For a borrower looking at a due-date reminder, begin with a renewal button and current due date and do not disclose the intended interpretation. Present the repaired artifact with the same surrounding interface but an independent fictional case; recruit someone not coached by the earlier test where feasible. The working question is whether the person can distinguish “a request action does not equal a successful state transition” from the tempting claim that clicking request renewal always extends the loan. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare unprompted decisions, explanations and remaining errors rather than merely tallying favorable reactions. The expected observation is that the person states which date controls while the renewal is pending. If instead the answer implies that clicking request renewal always extends the loan, log the exact words and where the reader looked. The potential harm is late return based on an unapproved extension. Do not treat a fluent reformulation as proof of meaning; request, confirmation and effective change are three separate events. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to display the current deadline until confirmation arrives. A corrected phrase is not validated by the editor’s confidence. Preserve before-and-after versions, source approval, exact prompts, reader evidence and unresolved risks. Report that the tested failure diminished under these conditions, not that the translation is universally impossible to misunderstand. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Domain studio 10: Research questionnaire scale

A a respondent completing a translated survey encounters a survey item and five response options. The intended rule is: The response scale asks about frequency during the past seven days, not intensity or lifetime experience. The risky alternative is a frequency option is interpreted as severity. This matters because unreliable comparisons across language versions. Here the semantic hinge is that measurement constructs can drift while phrasing sounds fluent. These are invented teaching fixtures, so a real publisher must replace them with an approved source contract and its own audience research.

Diagnostic case 082: Research questionnaire scale × Blind paraphrase

Reader-facing test. For a respondent completing a translated survey, begin with a survey item and five response options and do not disclose the intended interpretation. Withhold the source original and ask the reader, in their own language, to explain the message as if describing it to someone who has not seen the screen. The working question is whether the person can distinguish “measurement constructs can drift while phrasing sounds fluent” from the tempting claim that a frequency option is interpreted as severity. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Listen for the causal story: what triggers the action, what stays true afterward and which actor controls the change. The expected observation is that the person distinguishes how often from how strongly a feeling occurred. If instead the answer implies that a frequency option is interpreted as severity, log the exact words and where the reader looked. The potential harm is unreliable comparisons across language versions. Do not treat a fluent reformulation as proof of meaning; a plausible translation can change the variable being measured. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to clarify the reference period and response dimension without adding content. A clean paraphrase should preserve the decision rule rather than repeat isolated keywords. A polished but incorrect summary is more informative than a low-confidence guess because it shows the exact rule the reader has constructed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 083: Research questionnaire scale × Task simulation

Reader-facing test. For a respondent completing a translated survey, begin with a survey item and five response options and do not disclose the intended interpretation. Give a fictional scenario with a choice between acting now, waiting for confirmation and searching for another condition; allow the reader to use only the visible translated material. The working question is whether the person can distinguish “measurement constructs can drift while phrasing sounds fluent” from the tempting claim that a frequency option is interpreted as severity. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Capture the chosen action, the text used to justify it and whether a wrong branch seems attractive. The expected observation is that the person distinguishes how often from how strongly a feeling occurred. If instead the answer implies that a frequency option is interpreted as severity, log the exact words and where the reader looked. The potential harm is unreliable comparisons across language versions. Do not treat a fluent reformulation as proof of meaning; a plausible translation can change the variable being measured. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to clarify the reference period and response dimension without adding content. Do not score the final click alone. Someone may guess correctly from placement while holding the wrong belief. A complete result includes the reason, expected consequence and what additional information the person believes is needed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 084: Research questionnaire scale × Think-aloud trace

Reader-facing test. For a respondent completing a translated survey, begin with a survey item and five response options and do not disclose the intended interpretation. Ask the reader to narrate thoughts while moving from the notice to the relevant action, but do not correct terminology or explain the organization’s policy. The working question is whether the person can distinguish “measurement constructs can drift while phrasing sounds fluent” from the tempting claim that a frequency option is interpreted as severity. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Mark the moment a term is interpreted, the moment confidence changes and the point at which layout or icons redirect attention. The expected observation is that the person distinguishes how often from how strongly a feeling occurred. If instead the answer implies that a frequency option is interpreted as severity, log the exact words and where the reader looked. The potential harm is unreliable comparisons across language versions. Do not treat a fluent reformulation as proof of meaning; a plausible translation can change the variable being measured. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to clarify the reference period and response dimension without adding content. This method can identify an interface cue rather than a translation error. The very act of thinking aloud may change behavior, so report it as qualitative diagnostic evidence rather than a naturalistic time-to-completion measurement. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 085: Research questionnaire scale × Contrastive boundary pair

Reader-facing test. For a respondent completing a translated survey, begin with a survey item and five response options and do not disclose the intended interpretation. Present two nearly identical fictional cases that differ only in the decisive condition; ask whether the outcome changes and why. The working question is whether the person can distinguish “measurement constructs can drift while phrasing sounds fluent” from the tempting claim that a frequency option is interpreted as severity. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Look for the reader’s treatment of a status, exception or timing boundary that the source intended to preserve. The expected observation is that the person distinguishes how often from how strongly a feeling occurred. If instead the answer implies that a frequency option is interpreted as severity, log the exact words and where the reader looked. The potential harm is unreliable comparisons across language versions. Do not treat a fluent reformulation as proof of meaning; a plausible translation can change the variable being measured. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to clarify the reference period and response dimension without adding content. A reader who gives the same answer to both cases may have lost the qualifier. However, if the source rule itself is ambiguous, pause the test and clarify it with the content owner. Do not force respondents to guess at a rule the author has not specified. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 086: Research questionnaire scale × Actor and responsibility map

Reader-facing test. For a respondent completing a translated survey, begin with a survey item and five response options and do not disclose the intended interpretation. Ask the reader to identify who submits, who decides, who can revise the decision and who is affected; let them point to the translated words supporting each answer. The working question is whether the person can distinguish “measurement constructs can drift while phrasing sounds fluent” from the tempting claim that a frequency option is interpreted as severity. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare the named actors with the interpretation contract and note any responsibility the reader invents. The expected observation is that the person distinguishes how often from how strongly a feeling occurred. If instead the answer implies that a frequency option is interpreted as severity, log the exact words and where the reader looked. The potential harm is unreliable comparisons across language versions. Do not treat a fluent reformulation as proof of meaning; a plausible translation can change the variable being measured. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to clarify the reference period and response dimension without adding content. Passive voice is not automatically wrong, and some languages omit subjects naturally. The test is whether the actual actor is recoverable. If the reader supplies an unintended actor, identify whether nearby labels, prior notifications or grammatical choices created that inference. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 087: Research questionnaire scale × Counterexample challenge

Reader-facing test. For a respondent completing a translated survey, begin with a survey item and five response options and do not disclose the intended interpretation. Invite the reader to invent an example in which the notice would not apply, or to say what evidence would make the current instruction no longer valid. The working question is whether the person can distinguish “measurement constructs can drift while phrasing sounds fluent” from the tempting claim that a frequency option is interpreted as severity. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Inspect whether the reader recognizes limits, exceptions, unknown information and provisional states. The expected observation is that the person distinguishes how often from how strongly a feeling occurred. If instead the answer implies that a frequency option is interpreted as severity, log the exact words and where the reader looked. The potential harm is unreliable comparisons across language versions. Do not treat a fluent reformulation as proof of meaning; a plausible translation can change the variable being measured. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to clarify the reference period and response dimension without adding content. An inability to imagine a counterexample may be a testing skill issue, so offer a neutral hypothetical rather than marking immediate failure. The key evidence is the reader’s understanding of the boundary after an ordinary scenario is supplied without revealing the answer. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 088: Research questionnaire scale × Delayed teach-back

Reader-facing test. For a respondent completing a translated survey, begin with a survey item and five response options and do not disclose the intended interpretation. After an unrelated short filler activity, ask the reader to explain the next step to a fictional colleague who cannot view the translated message. The working question is whether the person can distinguish “measurement constructs can drift while phrasing sounds fluent” from the tempting claim that a frequency option is interpreted as severity. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Note which conditions survive memory, which qualifiers disappear and whether confidence rises as detail is lost. The expected observation is that the person distinguishes how often from how strongly a feeling occurred. If instead the answer implies that a frequency option is interpreted as severity, log the exact words and where the reader looked. The potential harm is unreliable comparisons across language versions. Do not treat a fluent reformulation as proof of meaning; a plausible translation can change the variable being measured. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to clarify the reference period and response dimension without adding content. Teach-back tests retained practical meaning, not memorization of exact wording. Compare it with first-exposure paraphrase. If the reader initially understood but later drops a critical exception, consider information placement, a reminder, a concise summary or a clearer call to action. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 089: Research questionnaire scale × Alternative wording comparison

Reader-facing test. For a respondent completing a translated survey, begin with a survey item and five response options and do not disclose the intended interpretation. Show two approved candidate renderings in a balanced order across participants and ask what each would lead them to do; avoid asking which sounds more elegant. The working question is whether the person can distinguish “measurement constructs can drift while phrasing sounds fluent” from the tempting claim that a frequency option is interpreted as severity. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Record whether the alternatives produce different real-world predictions despite similar fluency judgments. The expected observation is that the person distinguishes how often from how strongly a feeling occurred. If instead the answer implies that a frequency option is interpreted as severity, log the exact words and where the reader looked. The potential harm is unreliable comparisons across language versions. Do not treat a fluent reformulation as proof of meaning; a plausible translation can change the variable being measured. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to clarify the reference period and response dimension without adding content. Do not optimize language by majority preference alone: a popular label can still induce a serious wrong decision. Preference is a supporting measure. Select wording that better preserves the rule, is natural for the intended audience and survives a subsequent task-based retest. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 090: Research questionnaire scale × Release-readiness retest

Reader-facing test. For a respondent completing a translated survey, begin with a survey item and five response options and do not disclose the intended interpretation. Present the repaired artifact with the same surrounding interface but an independent fictional case; recruit someone not coached by the earlier test where feasible. The working question is whether the person can distinguish “measurement constructs can drift while phrasing sounds fluent” from the tempting claim that a frequency option is interpreted as severity. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare unprompted decisions, explanations and remaining errors rather than merely tallying favorable reactions. The expected observation is that the person distinguishes how often from how strongly a feeling occurred. If instead the answer implies that a frequency option is interpreted as severity, log the exact words and where the reader looked. The potential harm is unreliable comparisons across language versions. Do not treat a fluent reformulation as proof of meaning; a plausible translation can change the variable being measured. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to clarify the reference period and response dimension without adding content. A corrected phrase is not validated by the editor’s confidence. Preserve before-and-after versions, source approval, exact prompts, reader evidence and unresolved risks. Report that the tested failure diminished under these conditions, not that the translation is universally impossible to misunderstand. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Domain studio 11: Job-application status

A an applicant viewing a hiring portal encounters a status timeline and notification email. The intended rule is: A status of ‘received’ confirms submission only and does not imply human review, eligibility or an interview. The risky alternative is received means shortlisted. This matters because misleading expectations and missed required steps. Here the semantic hinge is that a transport acknowledgement is mistaken for a decision. These are invented teaching fixtures, so a real publisher must replace them with an approved source contract and its own audience research.

Diagnostic case 091: Job-application status × Blind paraphrase

Reader-facing test. For an applicant viewing a hiring portal, begin with a status timeline and notification email and do not disclose the intended interpretation. Withhold the source original and ask the reader, in their own language, to explain the message as if describing it to someone who has not seen the screen. The working question is whether the person can distinguish “a transport acknowledgement is mistaken for a decision” from the tempting claim that received means shortlisted. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Listen for the causal story: what triggers the action, what stays true afterward and which actor controls the change. The expected observation is that the person can name the next decision without claiming it has happened. If instead the answer implies that received means shortlisted, log the exact words and where the reader looked. The potential harm is misleading expectations and missed required steps. Do not treat a fluent reformulation as proof of meaning; workflow labels often inherit unearned promises. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to add a plain explanation of the precise submission stage. A clean paraphrase should preserve the decision rule rather than repeat isolated keywords. A polished but incorrect summary is more informative than a low-confidence guess because it shows the exact rule the reader has constructed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 092: Job-application status × Task simulation

Reader-facing test. For an applicant viewing a hiring portal, begin with a status timeline and notification email and do not disclose the intended interpretation. Give a fictional scenario with a choice between acting now, waiting for confirmation and searching for another condition; allow the reader to use only the visible translated material. The working question is whether the person can distinguish “a transport acknowledgement is mistaken for a decision” from the tempting claim that received means shortlisted. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Capture the chosen action, the text used to justify it and whether a wrong branch seems attractive. The expected observation is that the person can name the next decision without claiming it has happened. If instead the answer implies that received means shortlisted, log the exact words and where the reader looked. The potential harm is misleading expectations and missed required steps. Do not treat a fluent reformulation as proof of meaning; workflow labels often inherit unearned promises. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to add a plain explanation of the precise submission stage. Do not score the final click alone. Someone may guess correctly from placement while holding the wrong belief. A complete result includes the reason, expected consequence and what additional information the person believes is needed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 093: Job-application status × Think-aloud trace

Reader-facing test. For an applicant viewing a hiring portal, begin with a status timeline and notification email and do not disclose the intended interpretation. Ask the reader to narrate thoughts while moving from the notice to the relevant action, but do not correct terminology or explain the organization’s policy. The working question is whether the person can distinguish “a transport acknowledgement is mistaken for a decision” from the tempting claim that received means shortlisted. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Mark the moment a term is interpreted, the moment confidence changes and the point at which layout or icons redirect attention. The expected observation is that the person can name the next decision without claiming it has happened. If instead the answer implies that received means shortlisted, log the exact words and where the reader looked. The potential harm is misleading expectations and missed required steps. Do not treat a fluent reformulation as proof of meaning; workflow labels often inherit unearned promises. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to add a plain explanation of the precise submission stage. This method can identify an interface cue rather than a translation error. The very act of thinking aloud may change behavior, so report it as qualitative diagnostic evidence rather than a naturalistic time-to-completion measurement. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 094: Job-application status × Contrastive boundary pair

Reader-facing test. For an applicant viewing a hiring portal, begin with a status timeline and notification email and do not disclose the intended interpretation. Present two nearly identical fictional cases that differ only in the decisive condition; ask whether the outcome changes and why. The working question is whether the person can distinguish “a transport acknowledgement is mistaken for a decision” from the tempting claim that received means shortlisted. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Look for the reader’s treatment of a status, exception or timing boundary that the source intended to preserve. The expected observation is that the person can name the next decision without claiming it has happened. If instead the answer implies that received means shortlisted, log the exact words and where the reader looked. The potential harm is misleading expectations and missed required steps. Do not treat a fluent reformulation as proof of meaning; workflow labels often inherit unearned promises. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to add a plain explanation of the precise submission stage. A reader who gives the same answer to both cases may have lost the qualifier. However, if the source rule itself is ambiguous, pause the test and clarify it with the content owner. Do not force respondents to guess at a rule the author has not specified. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 095: Job-application status × Actor and responsibility map

Reader-facing test. For an applicant viewing a hiring portal, begin with a status timeline and notification email and do not disclose the intended interpretation. Ask the reader to identify who submits, who decides, who can revise the decision and who is affected; let them point to the translated words supporting each answer. The working question is whether the person can distinguish “a transport acknowledgement is mistaken for a decision” from the tempting claim that received means shortlisted. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare the named actors with the interpretation contract and note any responsibility the reader invents. The expected observation is that the person can name the next decision without claiming it has happened. If instead the answer implies that received means shortlisted, log the exact words and where the reader looked. The potential harm is misleading expectations and missed required steps. Do not treat a fluent reformulation as proof of meaning; workflow labels often inherit unearned promises. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to add a plain explanation of the precise submission stage. Passive voice is not automatically wrong, and some languages omit subjects naturally. The test is whether the actual actor is recoverable. If the reader supplies an unintended actor, identify whether nearby labels, prior notifications or grammatical choices created that inference. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 096: Job-application status × Counterexample challenge

Reader-facing test. For an applicant viewing a hiring portal, begin with a status timeline and notification email and do not disclose the intended interpretation. Invite the reader to invent an example in which the notice would not apply, or to say what evidence would make the current instruction no longer valid. The working question is whether the person can distinguish “a transport acknowledgement is mistaken for a decision” from the tempting claim that received means shortlisted. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Inspect whether the reader recognizes limits, exceptions, unknown information and provisional states. The expected observation is that the person can name the next decision without claiming it has happened. If instead the answer implies that received means shortlisted, log the exact words and where the reader looked. The potential harm is misleading expectations and missed required steps. Do not treat a fluent reformulation as proof of meaning; workflow labels often inherit unearned promises. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to add a plain explanation of the precise submission stage. An inability to imagine a counterexample may be a testing skill issue, so offer a neutral hypothetical rather than marking immediate failure. The key evidence is the reader’s understanding of the boundary after an ordinary scenario is supplied without revealing the answer. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 097: Job-application status × Delayed teach-back

Reader-facing test. For an applicant viewing a hiring portal, begin with a status timeline and notification email and do not disclose the intended interpretation. After an unrelated short filler activity, ask the reader to explain the next step to a fictional colleague who cannot view the translated message. The working question is whether the person can distinguish “a transport acknowledgement is mistaken for a decision” from the tempting claim that received means shortlisted. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Note which conditions survive memory, which qualifiers disappear and whether confidence rises as detail is lost. The expected observation is that the person can name the next decision without claiming it has happened. If instead the answer implies that received means shortlisted, log the exact words and where the reader looked. The potential harm is misleading expectations and missed required steps. Do not treat a fluent reformulation as proof of meaning; workflow labels often inherit unearned promises. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to add a plain explanation of the precise submission stage. Teach-back tests retained practical meaning, not memorization of exact wording. Compare it with first-exposure paraphrase. If the reader initially understood but later drops a critical exception, consider information placement, a reminder, a concise summary or a clearer call to action. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 098: Job-application status × Alternative wording comparison

Reader-facing test. For an applicant viewing a hiring portal, begin with a status timeline and notification email and do not disclose the intended interpretation. Show two approved candidate renderings in a balanced order across participants and ask what each would lead them to do; avoid asking which sounds more elegant. The working question is whether the person can distinguish “a transport acknowledgement is mistaken for a decision” from the tempting claim that received means shortlisted. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Record whether the alternatives produce different real-world predictions despite similar fluency judgments. The expected observation is that the person can name the next decision without claiming it has happened. If instead the answer implies that received means shortlisted, log the exact words and where the reader looked. The potential harm is misleading expectations and missed required steps. Do not treat a fluent reformulation as proof of meaning; workflow labels often inherit unearned promises. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to add a plain explanation of the precise submission stage. Do not optimize language by majority preference alone: a popular label can still induce a serious wrong decision. Preference is a supporting measure. Select wording that better preserves the rule, is natural for the intended audience and survives a subsequent task-based retest. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 099: Job-application status × Release-readiness retest

Reader-facing test. For an applicant viewing a hiring portal, begin with a status timeline and notification email and do not disclose the intended interpretation. Present the repaired artifact with the same surrounding interface but an independent fictional case; recruit someone not coached by the earlier test where feasible. The working question is whether the person can distinguish “a transport acknowledgement is mistaken for a decision” from the tempting claim that received means shortlisted. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare unprompted decisions, explanations and remaining errors rather than merely tallying favorable reactions. The expected observation is that the person can name the next decision without claiming it has happened. If instead the answer implies that received means shortlisted, log the exact words and where the reader looked. The potential harm is misleading expectations and missed required steps. Do not treat a fluent reformulation as proof of meaning; workflow labels often inherit unearned promises. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to add a plain explanation of the precise submission stage. A corrected phrase is not validated by the editor’s confidence. Preserve before-and-after versions, source approval, exact prompts, reader evidence and unresolved risks. Report that the tested failure diminished under these conditions, not that the translation is universally impossible to misunderstand. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Domain studio 12: Remote-work shift roster

A a team member checking schedule changes encounters a chat-like shift request and approval badge. The intended rule is: A proposed shift swap remains tentative until the designated approver confirms it; both people remain bound by the current roster meanwhile. The risky alternative is a colleague’s agreement finalizes the roster. This matters because uncovered duties and avoidable scheduling conflict. Here the semantic hinge is that social agreement is not scheduling authorization. These are invented teaching fixtures, so a real publisher must replace them with an approved source contract and its own audience research.

Diagnostic case 100: Remote-work shift roster × Blind paraphrase

Reader-facing test. For a team member checking schedule changes, begin with a chat-like shift request and approval badge and do not disclose the intended interpretation. Withhold the source original and ask the reader, in their own language, to explain the message as if describing it to someone who has not seen the screen. The working question is whether the person can distinguish “social agreement is not scheduling authorization” from the tempting claim that a colleague’s agreement finalizes the roster. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Listen for the causal story: what triggers the action, what stays true afterward and which actor controls the change. The expected observation is that the person identifies whose approval changes the work obligation. If instead the answer implies that a colleague’s agreement finalizes the roster, log the exact words and where the reader looked. The potential harm is uncovered duties and avoidable scheduling conflict. Do not treat a fluent reformulation as proof of meaning; a conversational ‘yes’ is not always the final permission. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to distinguish proposal, peer agreement and final authorization. A clean paraphrase should preserve the decision rule rather than repeat isolated keywords. A polished but incorrect summary is more informative than a low-confidence guess because it shows the exact rule the reader has constructed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 101: Remote-work shift roster × Task simulation

Reader-facing test. For a team member checking schedule changes, begin with a chat-like shift request and approval badge and do not disclose the intended interpretation. Give a fictional scenario with a choice between acting now, waiting for confirmation and searching for another condition; allow the reader to use only the visible translated material. The working question is whether the person can distinguish “social agreement is not scheduling authorization” from the tempting claim that a colleague’s agreement finalizes the roster. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Capture the chosen action, the text used to justify it and whether a wrong branch seems attractive. The expected observation is that the person identifies whose approval changes the work obligation. If instead the answer implies that a colleague’s agreement finalizes the roster, log the exact words and where the reader looked. The potential harm is uncovered duties and avoidable scheduling conflict. Do not treat a fluent reformulation as proof of meaning; a conversational ‘yes’ is not always the final permission. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to distinguish proposal, peer agreement and final authorization. Do not score the final click alone. Someone may guess correctly from placement while holding the wrong belief. A complete result includes the reason, expected consequence and what additional information the person believes is needed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 102: Remote-work shift roster × Think-aloud trace

Reader-facing test. For a team member checking schedule changes, begin with a chat-like shift request and approval badge and do not disclose the intended interpretation. Ask the reader to narrate thoughts while moving from the notice to the relevant action, but do not correct terminology or explain the organization’s policy. The working question is whether the person can distinguish “social agreement is not scheduling authorization” from the tempting claim that a colleague’s agreement finalizes the roster. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Mark the moment a term is interpreted, the moment confidence changes and the point at which layout or icons redirect attention. The expected observation is that the person identifies whose approval changes the work obligation. If instead the answer implies that a colleague’s agreement finalizes the roster, log the exact words and where the reader looked. The potential harm is uncovered duties and avoidable scheduling conflict. Do not treat a fluent reformulation as proof of meaning; a conversational ‘yes’ is not always the final permission. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to distinguish proposal, peer agreement and final authorization. This method can identify an interface cue rather than a translation error. The very act of thinking aloud may change behavior, so report it as qualitative diagnostic evidence rather than a naturalistic time-to-completion measurement. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 103: Remote-work shift roster × Contrastive boundary pair

Reader-facing test. For a team member checking schedule changes, begin with a chat-like shift request and approval badge and do not disclose the intended interpretation. Present two nearly identical fictional cases that differ only in the decisive condition; ask whether the outcome changes and why. The working question is whether the person can distinguish “social agreement is not scheduling authorization” from the tempting claim that a colleague’s agreement finalizes the roster. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Look for the reader’s treatment of a status, exception or timing boundary that the source intended to preserve. The expected observation is that the person identifies whose approval changes the work obligation. If instead the answer implies that a colleague’s agreement finalizes the roster, log the exact words and where the reader looked. The potential harm is uncovered duties and avoidable scheduling conflict. Do not treat a fluent reformulation as proof of meaning; a conversational ‘yes’ is not always the final permission. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to distinguish proposal, peer agreement and final authorization. A reader who gives the same answer to both cases may have lost the qualifier. However, if the source rule itself is ambiguous, pause the test and clarify it with the content owner. Do not force respondents to guess at a rule the author has not specified. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 104: Remote-work shift roster × Actor and responsibility map

Reader-facing test. For a team member checking schedule changes, begin with a chat-like shift request and approval badge and do not disclose the intended interpretation. Ask the reader to identify who submits, who decides, who can revise the decision and who is affected; let them point to the translated words supporting each answer. The working question is whether the person can distinguish “social agreement is not scheduling authorization” from the tempting claim that a colleague’s agreement finalizes the roster. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare the named actors with the interpretation contract and note any responsibility the reader invents. The expected observation is that the person identifies whose approval changes the work obligation. If instead the answer implies that a colleague’s agreement finalizes the roster, log the exact words and where the reader looked. The potential harm is uncovered duties and avoidable scheduling conflict. Do not treat a fluent reformulation as proof of meaning; a conversational ‘yes’ is not always the final permission. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to distinguish proposal, peer agreement and final authorization. Passive voice is not automatically wrong, and some languages omit subjects naturally. The test is whether the actual actor is recoverable. If the reader supplies an unintended actor, identify whether nearby labels, prior notifications or grammatical choices created that inference. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 105: Remote-work shift roster × Counterexample challenge

Reader-facing test. For a team member checking schedule changes, begin with a chat-like shift request and approval badge and do not disclose the intended interpretation. Invite the reader to invent an example in which the notice would not apply, or to say what evidence would make the current instruction no longer valid. The working question is whether the person can distinguish “social agreement is not scheduling authorization” from the tempting claim that a colleague’s agreement finalizes the roster. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Inspect whether the reader recognizes limits, exceptions, unknown information and provisional states. The expected observation is that the person identifies whose approval changes the work obligation. If instead the answer implies that a colleague’s agreement finalizes the roster, log the exact words and where the reader looked. The potential harm is uncovered duties and avoidable scheduling conflict. Do not treat a fluent reformulation as proof of meaning; a conversational ‘yes’ is not always the final permission. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to distinguish proposal, peer agreement and final authorization. An inability to imagine a counterexample may be a testing skill issue, so offer a neutral hypothetical rather than marking immediate failure. The key evidence is the reader’s understanding of the boundary after an ordinary scenario is supplied without revealing the answer. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 106: Remote-work shift roster × Delayed teach-back

Reader-facing test. For a team member checking schedule changes, begin with a chat-like shift request and approval badge and do not disclose the intended interpretation. After an unrelated short filler activity, ask the reader to explain the next step to a fictional colleague who cannot view the translated message. The working question is whether the person can distinguish “social agreement is not scheduling authorization” from the tempting claim that a colleague’s agreement finalizes the roster. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Note which conditions survive memory, which qualifiers disappear and whether confidence rises as detail is lost. The expected observation is that the person identifies whose approval changes the work obligation. If instead the answer implies that a colleague’s agreement finalizes the roster, log the exact words and where the reader looked. The potential harm is uncovered duties and avoidable scheduling conflict. Do not treat a fluent reformulation as proof of meaning; a conversational ‘yes’ is not always the final permission. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to distinguish proposal, peer agreement and final authorization. Teach-back tests retained practical meaning, not memorization of exact wording. Compare it with first-exposure paraphrase. If the reader initially understood but later drops a critical exception, consider information placement, a reminder, a concise summary or a clearer call to action. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 107: Remote-work shift roster × Alternative wording comparison

Reader-facing test. For a team member checking schedule changes, begin with a chat-like shift request and approval badge and do not disclose the intended interpretation. Show two approved candidate renderings in a balanced order across participants and ask what each would lead them to do; avoid asking which sounds more elegant. The working question is whether the person can distinguish “social agreement is not scheduling authorization” from the tempting claim that a colleague’s agreement finalizes the roster. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Record whether the alternatives produce different real-world predictions despite similar fluency judgments. The expected observation is that the person identifies whose approval changes the work obligation. If instead the answer implies that a colleague’s agreement finalizes the roster, log the exact words and where the reader looked. The potential harm is uncovered duties and avoidable scheduling conflict. Do not treat a fluent reformulation as proof of meaning; a conversational ‘yes’ is not always the final permission. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to distinguish proposal, peer agreement and final authorization. Do not optimize language by majority preference alone: a popular label can still induce a serious wrong decision. Preference is a supporting measure. Select wording that better preserves the rule, is natural for the intended audience and survives a subsequent task-based retest. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 108: Remote-work shift roster × Release-readiness retest

Reader-facing test. For a team member checking schedule changes, begin with a chat-like shift request and approval badge and do not disclose the intended interpretation. Present the repaired artifact with the same surrounding interface but an independent fictional case; recruit someone not coached by the earlier test where feasible. The working question is whether the person can distinguish “social agreement is not scheduling authorization” from the tempting claim that a colleague’s agreement finalizes the roster. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare unprompted decisions, explanations and remaining errors rather than merely tallying favorable reactions. The expected observation is that the person identifies whose approval changes the work obligation. If instead the answer implies that a colleague’s agreement finalizes the roster, log the exact words and where the reader looked. The potential harm is uncovered duties and avoidable scheduling conflict. Do not treat a fluent reformulation as proof of meaning; a conversational ‘yes’ is not always the final permission. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to distinguish proposal, peer agreement and final authorization. A corrected phrase is not validated by the editor’s confidence. Preserve before-and-after versions, source approval, exact prompts, reader evidence and unresolved risks. Report that the tested failure diminished under these conditions, not that the translation is universally impossible to misunderstand. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Domain studio 13: International delivery cost

A a buyer reviewing an online order encounters an order summary with an estimated total. The intended rule is: The displayed product price excludes any charges that the checkout explicitly identifies as separately assessed; estimate is not final invoice. The risky alternative is an estimate is a binding total. This matters because surprise charges and ill-informed purchase decisions. Here the semantic hinge is that a preliminary number is confused with a final amount. These are invented teaching fixtures, so a real publisher must replace them with an approved source contract and its own audience research.

Diagnostic case 109: International delivery cost × Blind paraphrase

Reader-facing test. For a buyer reviewing an online order, begin with an order summary with an estimated total and do not disclose the intended interpretation. Withhold the source original and ask the reader, in their own language, to explain the message as if describing it to someone who has not seen the screen. The working question is whether the person can distinguish “a preliminary number is confused with a final amount” from the tempting claim that an estimate is a binding total. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Listen for the causal story: what triggers the action, what stays true afterward and which actor controls the change. The expected observation is that the person can say what is known now and what remains conditional. If instead the answer implies that an estimate is a binding total, log the exact words and where the reader looked. The potential harm is surprise charges and ill-informed purchase decisions. Do not treat a fluent reformulation as proof of meaning; certainty words matter as much as currency symbols. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label estimated and confirmed amounts using consistent qualifiers. A clean paraphrase should preserve the decision rule rather than repeat isolated keywords. A polished but incorrect summary is more informative than a low-confidence guess because it shows the exact rule the reader has constructed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 110: International delivery cost × Task simulation

Reader-facing test. For a buyer reviewing an online order, begin with an order summary with an estimated total and do not disclose the intended interpretation. Give a fictional scenario with a choice between acting now, waiting for confirmation and searching for another condition; allow the reader to use only the visible translated material. The working question is whether the person can distinguish “a preliminary number is confused with a final amount” from the tempting claim that an estimate is a binding total. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Capture the chosen action, the text used to justify it and whether a wrong branch seems attractive. The expected observation is that the person can say what is known now and what remains conditional. If instead the answer implies that an estimate is a binding total, log the exact words and where the reader looked. The potential harm is surprise charges and ill-informed purchase decisions. Do not treat a fluent reformulation as proof of meaning; certainty words matter as much as currency symbols. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label estimated and confirmed amounts using consistent qualifiers. Do not score the final click alone. Someone may guess correctly from placement while holding the wrong belief. A complete result includes the reason, expected consequence and what additional information the person believes is needed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 111: International delivery cost × Think-aloud trace

Reader-facing test. For a buyer reviewing an online order, begin with an order summary with an estimated total and do not disclose the intended interpretation. Ask the reader to narrate thoughts while moving from the notice to the relevant action, but do not correct terminology or explain the organization’s policy. The working question is whether the person can distinguish “a preliminary number is confused with a final amount” from the tempting claim that an estimate is a binding total. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Mark the moment a term is interpreted, the moment confidence changes and the point at which layout or icons redirect attention. The expected observation is that the person can say what is known now and what remains conditional. If instead the answer implies that an estimate is a binding total, log the exact words and where the reader looked. The potential harm is surprise charges and ill-informed purchase decisions. Do not treat a fluent reformulation as proof of meaning; certainty words matter as much as currency symbols. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label estimated and confirmed amounts using consistent qualifiers. This method can identify an interface cue rather than a translation error. The very act of thinking aloud may change behavior, so report it as qualitative diagnostic evidence rather than a naturalistic time-to-completion measurement. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 112: International delivery cost × Contrastive boundary pair

Reader-facing test. For a buyer reviewing an online order, begin with an order summary with an estimated total and do not disclose the intended interpretation. Present two nearly identical fictional cases that differ only in the decisive condition; ask whether the outcome changes and why. The working question is whether the person can distinguish “a preliminary number is confused with a final amount” from the tempting claim that an estimate is a binding total. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Look for the reader’s treatment of a status, exception or timing boundary that the source intended to preserve. The expected observation is that the person can say what is known now and what remains conditional. If instead the answer implies that an estimate is a binding total, log the exact words and where the reader looked. The potential harm is surprise charges and ill-informed purchase decisions. Do not treat a fluent reformulation as proof of meaning; certainty words matter as much as currency symbols. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label estimated and confirmed amounts using consistent qualifiers. A reader who gives the same answer to both cases may have lost the qualifier. However, if the source rule itself is ambiguous, pause the test and clarify it with the content owner. Do not force respondents to guess at a rule the author has not specified. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 113: International delivery cost × Actor and responsibility map

Reader-facing test. For a buyer reviewing an online order, begin with an order summary with an estimated total and do not disclose the intended interpretation. Ask the reader to identify who submits, who decides, who can revise the decision and who is affected; let them point to the translated words supporting each answer. The working question is whether the person can distinguish “a preliminary number is confused with a final amount” from the tempting claim that an estimate is a binding total. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare the named actors with the interpretation contract and note any responsibility the reader invents. The expected observation is that the person can say what is known now and what remains conditional. If instead the answer implies that an estimate is a binding total, log the exact words and where the reader looked. The potential harm is surprise charges and ill-informed purchase decisions. Do not treat a fluent reformulation as proof of meaning; certainty words matter as much as currency symbols. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label estimated and confirmed amounts using consistent qualifiers. Passive voice is not automatically wrong, and some languages omit subjects naturally. The test is whether the actual actor is recoverable. If the reader supplies an unintended actor, identify whether nearby labels, prior notifications or grammatical choices created that inference. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 114: International delivery cost × Counterexample challenge

Reader-facing test. For a buyer reviewing an online order, begin with an order summary with an estimated total and do not disclose the intended interpretation. Invite the reader to invent an example in which the notice would not apply, or to say what evidence would make the current instruction no longer valid. The working question is whether the person can distinguish “a preliminary number is confused with a final amount” from the tempting claim that an estimate is a binding total. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Inspect whether the reader recognizes limits, exceptions, unknown information and provisional states. The expected observation is that the person can say what is known now and what remains conditional. If instead the answer implies that an estimate is a binding total, log the exact words and where the reader looked. The potential harm is surprise charges and ill-informed purchase decisions. Do not treat a fluent reformulation as proof of meaning; certainty words matter as much as currency symbols. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label estimated and confirmed amounts using consistent qualifiers. An inability to imagine a counterexample may be a testing skill issue, so offer a neutral hypothetical rather than marking immediate failure. The key evidence is the reader’s understanding of the boundary after an ordinary scenario is supplied without revealing the answer. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 115: International delivery cost × Delayed teach-back

Reader-facing test. For a buyer reviewing an online order, begin with an order summary with an estimated total and do not disclose the intended interpretation. After an unrelated short filler activity, ask the reader to explain the next step to a fictional colleague who cannot view the translated message. The working question is whether the person can distinguish “a preliminary number is confused with a final amount” from the tempting claim that an estimate is a binding total. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Note which conditions survive memory, which qualifiers disappear and whether confidence rises as detail is lost. The expected observation is that the person can say what is known now and what remains conditional. If instead the answer implies that an estimate is a binding total, log the exact words and where the reader looked. The potential harm is surprise charges and ill-informed purchase decisions. Do not treat a fluent reformulation as proof of meaning; certainty words matter as much as currency symbols. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label estimated and confirmed amounts using consistent qualifiers. Teach-back tests retained practical meaning, not memorization of exact wording. Compare it with first-exposure paraphrase. If the reader initially understood but later drops a critical exception, consider information placement, a reminder, a concise summary or a clearer call to action. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 116: International delivery cost × Alternative wording comparison

Reader-facing test. For a buyer reviewing an online order, begin with an order summary with an estimated total and do not disclose the intended interpretation. Show two approved candidate renderings in a balanced order across participants and ask what each would lead them to do; avoid asking which sounds more elegant. The working question is whether the person can distinguish “a preliminary number is confused with a final amount” from the tempting claim that an estimate is a binding total. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Record whether the alternatives produce different real-world predictions despite similar fluency judgments. The expected observation is that the person can say what is known now and what remains conditional. If instead the answer implies that an estimate is a binding total, log the exact words and where the reader looked. The potential harm is surprise charges and ill-informed purchase decisions. Do not treat a fluent reformulation as proof of meaning; certainty words matter as much as currency symbols. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label estimated and confirmed amounts using consistent qualifiers. Do not optimize language by majority preference alone: a popular label can still induce a serious wrong decision. Preference is a supporting measure. Select wording that better preserves the rule, is natural for the intended audience and survives a subsequent task-based retest. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 117: International delivery cost × Release-readiness retest

Reader-facing test. For a buyer reviewing an online order, begin with an order summary with an estimated total and do not disclose the intended interpretation. Present the repaired artifact with the same surrounding interface but an independent fictional case; recruit someone not coached by the earlier test where feasible. The working question is whether the person can distinguish “a preliminary number is confused with a final amount” from the tempting claim that an estimate is a binding total. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare unprompted decisions, explanations and remaining errors rather than merely tallying favorable reactions. The expected observation is that the person can say what is known now and what remains conditional. If instead the answer implies that an estimate is a binding total, log the exact words and where the reader looked. The potential harm is surprise charges and ill-informed purchase decisions. Do not treat a fluent reformulation as proof of meaning; certainty words matter as much as currency symbols. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label estimated and confirmed amounts using consistent qualifiers. A corrected phrase is not validated by the editor’s confidence. Preserve before-and-after versions, source approval, exact prompts, reader evidence and unresolved risks. Report that the tested failure diminished under these conditions, not that the translation is universally impossible to misunderstand. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Domain studio 14: Child-account permission

A a guardian reviewing app settings encounters a toggle and adjacent description of data sharing. The intended rule is: Allowing a limited feature does not enable unrelated sharing or public visibility; the chosen permission applies only to the named feature. The risky alternative is enabling one option grants every related permission. This matters because unwanted disclosure or mistaken parental control. Here the semantic hinge is that scope boundaries disappear in a friendly-sounding label. These are invented teaching fixtures, so a real publisher must replace them with an approved source contract and its own audience research.

Diagnostic case 118: Child-account permission × Blind paraphrase

Reader-facing test. For a guardian reviewing app settings, begin with a toggle and adjacent description of data sharing and do not disclose the intended interpretation. Withhold the source original and ask the reader, in their own language, to explain the message as if describing it to someone who has not seen the screen. The working question is whether the person can distinguish “scope boundaries disappear in a friendly-sounding label” from the tempting claim that enabling one option grants every related permission. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Listen for the causal story: what triggers the action, what stays true afterward and which actor controls the change. The expected observation is that the person can tell which feature changes and which stays unchanged. If instead the answer implies that enabling one option grants every related permission, log the exact words and where the reader looked. The potential harm is unwanted disclosure or mistaken parental control. Do not treat a fluent reformulation as proof of meaning; an umbrella word like ‘access’ can hide multiple rights. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to name the allowed feature and unaffected settings explicitly. A clean paraphrase should preserve the decision rule rather than repeat isolated keywords. A polished but incorrect summary is more informative than a low-confidence guess because it shows the exact rule the reader has constructed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 119: Child-account permission × Task simulation

Reader-facing test. For a guardian reviewing app settings, begin with a toggle and adjacent description of data sharing and do not disclose the intended interpretation. Give a fictional scenario with a choice between acting now, waiting for confirmation and searching for another condition; allow the reader to use only the visible translated material. The working question is whether the person can distinguish “scope boundaries disappear in a friendly-sounding label” from the tempting claim that enabling one option grants every related permission. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Capture the chosen action, the text used to justify it and whether a wrong branch seems attractive. The expected observation is that the person can tell which feature changes and which stays unchanged. If instead the answer implies that enabling one option grants every related permission, log the exact words and where the reader looked. The potential harm is unwanted disclosure or mistaken parental control. Do not treat a fluent reformulation as proof of meaning; an umbrella word like ‘access’ can hide multiple rights. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to name the allowed feature and unaffected settings explicitly. Do not score the final click alone. Someone may guess correctly from placement while holding the wrong belief. A complete result includes the reason, expected consequence and what additional information the person believes is needed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 120: Child-account permission × Think-aloud trace

Reader-facing test. For a guardian reviewing app settings, begin with a toggle and adjacent description of data sharing and do not disclose the intended interpretation. Ask the reader to narrate thoughts while moving from the notice to the relevant action, but do not correct terminology or explain the organization’s policy. The working question is whether the person can distinguish “scope boundaries disappear in a friendly-sounding label” from the tempting claim that enabling one option grants every related permission. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Mark the moment a term is interpreted, the moment confidence changes and the point at which layout or icons redirect attention. The expected observation is that the person can tell which feature changes and which stays unchanged. If instead the answer implies that enabling one option grants every related permission, log the exact words and where the reader looked. The potential harm is unwanted disclosure or mistaken parental control. Do not treat a fluent reformulation as proof of meaning; an umbrella word like ‘access’ can hide multiple rights. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to name the allowed feature and unaffected settings explicitly. This method can identify an interface cue rather than a translation error. The very act of thinking aloud may change behavior, so report it as qualitative diagnostic evidence rather than a naturalistic time-to-completion measurement. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 121: Child-account permission × Contrastive boundary pair

Reader-facing test. For a guardian reviewing app settings, begin with a toggle and adjacent description of data sharing and do not disclose the intended interpretation. Present two nearly identical fictional cases that differ only in the decisive condition; ask whether the outcome changes and why. The working question is whether the person can distinguish “scope boundaries disappear in a friendly-sounding label” from the tempting claim that enabling one option grants every related permission. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Look for the reader’s treatment of a status, exception or timing boundary that the source intended to preserve. The expected observation is that the person can tell which feature changes and which stays unchanged. If instead the answer implies that enabling one option grants every related permission, log the exact words and where the reader looked. The potential harm is unwanted disclosure or mistaken parental control. Do not treat a fluent reformulation as proof of meaning; an umbrella word like ‘access’ can hide multiple rights. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to name the allowed feature and unaffected settings explicitly. A reader who gives the same answer to both cases may have lost the qualifier. However, if the source rule itself is ambiguous, pause the test and clarify it with the content owner. Do not force respondents to guess at a rule the author has not specified. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 122: Child-account permission × Actor and responsibility map

Reader-facing test. For a guardian reviewing app settings, begin with a toggle and adjacent description of data sharing and do not disclose the intended interpretation. Ask the reader to identify who submits, who decides, who can revise the decision and who is affected; let them point to the translated words supporting each answer. The working question is whether the person can distinguish “scope boundaries disappear in a friendly-sounding label” from the tempting claim that enabling one option grants every related permission. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare the named actors with the interpretation contract and note any responsibility the reader invents. The expected observation is that the person can tell which feature changes and which stays unchanged. If instead the answer implies that enabling one option grants every related permission, log the exact words and where the reader looked. The potential harm is unwanted disclosure or mistaken parental control. Do not treat a fluent reformulation as proof of meaning; an umbrella word like ‘access’ can hide multiple rights. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to name the allowed feature and unaffected settings explicitly. Passive voice is not automatically wrong, and some languages omit subjects naturally. The test is whether the actual actor is recoverable. If the reader supplies an unintended actor, identify whether nearby labels, prior notifications or grammatical choices created that inference. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 123: Child-account permission × Counterexample challenge

Reader-facing test. For a guardian reviewing app settings, begin with a toggle and adjacent description of data sharing and do not disclose the intended interpretation. Invite the reader to invent an example in which the notice would not apply, or to say what evidence would make the current instruction no longer valid. The working question is whether the person can distinguish “scope boundaries disappear in a friendly-sounding label” from the tempting claim that enabling one option grants every related permission. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Inspect whether the reader recognizes limits, exceptions, unknown information and provisional states. The expected observation is that the person can tell which feature changes and which stays unchanged. If instead the answer implies that enabling one option grants every related permission, log the exact words and where the reader looked. The potential harm is unwanted disclosure or mistaken parental control. Do not treat a fluent reformulation as proof of meaning; an umbrella word like ‘access’ can hide multiple rights. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to name the allowed feature and unaffected settings explicitly. An inability to imagine a counterexample may be a testing skill issue, so offer a neutral hypothetical rather than marking immediate failure. The key evidence is the reader’s understanding of the boundary after an ordinary scenario is supplied without revealing the answer. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 124: Child-account permission × Delayed teach-back

Reader-facing test. For a guardian reviewing app settings, begin with a toggle and adjacent description of data sharing and do not disclose the intended interpretation. After an unrelated short filler activity, ask the reader to explain the next step to a fictional colleague who cannot view the translated message. The working question is whether the person can distinguish “scope boundaries disappear in a friendly-sounding label” from the tempting claim that enabling one option grants every related permission. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Note which conditions survive memory, which qualifiers disappear and whether confidence rises as detail is lost. The expected observation is that the person can tell which feature changes and which stays unchanged. If instead the answer implies that enabling one option grants every related permission, log the exact words and where the reader looked. The potential harm is unwanted disclosure or mistaken parental control. Do not treat a fluent reformulation as proof of meaning; an umbrella word like ‘access’ can hide multiple rights. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to name the allowed feature and unaffected settings explicitly. Teach-back tests retained practical meaning, not memorization of exact wording. Compare it with first-exposure paraphrase. If the reader initially understood but later drops a critical exception, consider information placement, a reminder, a concise summary or a clearer call to action. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 125: Child-account permission × Alternative wording comparison

Reader-facing test. For a guardian reviewing app settings, begin with a toggle and adjacent description of data sharing and do not disclose the intended interpretation. Show two approved candidate renderings in a balanced order across participants and ask what each would lead them to do; avoid asking which sounds more elegant. The working question is whether the person can distinguish “scope boundaries disappear in a friendly-sounding label” from the tempting claim that enabling one option grants every related permission. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Record whether the alternatives produce different real-world predictions despite similar fluency judgments. The expected observation is that the person can tell which feature changes and which stays unchanged. If instead the answer implies that enabling one option grants every related permission, log the exact words and where the reader looked. The potential harm is unwanted disclosure or mistaken parental control. Do not treat a fluent reformulation as proof of meaning; an umbrella word like ‘access’ can hide multiple rights. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to name the allowed feature and unaffected settings explicitly. Do not optimize language by majority preference alone: a popular label can still induce a serious wrong decision. Preference is a supporting measure. Select wording that better preserves the rule, is natural for the intended audience and survives a subsequent task-based retest. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 126: Child-account permission × Release-readiness retest

Reader-facing test. For a guardian reviewing app settings, begin with a toggle and adjacent description of data sharing and do not disclose the intended interpretation. Present the repaired artifact with the same surrounding interface but an independent fictional case; recruit someone not coached by the earlier test where feasible. The working question is whether the person can distinguish “scope boundaries disappear in a friendly-sounding label” from the tempting claim that enabling one option grants every related permission. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare unprompted decisions, explanations and remaining errors rather than merely tallying favorable reactions. The expected observation is that the person can tell which feature changes and which stays unchanged. If instead the answer implies that enabling one option grants every related permission, log the exact words and where the reader looked. The potential harm is unwanted disclosure or mistaken parental control. Do not treat a fluent reformulation as proof of meaning; an umbrella word like ‘access’ can hide multiple rights. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to name the allowed feature and unaffected settings explicitly. A corrected phrase is not validated by the editor’s confidence. Preserve before-and-after versions, source approval, exact prompts, reader evidence and unresolved risks. Report that the tested failure diminished under these conditions, not that the translation is universally impossible to misunderstand. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Domain studio 15: Open-source license notice

A a developer reading redistribution guidance encounters a download page with a license link. The intended rule is: A permission to reuse material is subject to the named license conditions; attribution and redistribution duties are not optional merely because access is free. The risky alternative is free availability means no obligations. This matters because unintended redistribution or omission of required notices. Here the semantic hinge is that price, permission and compliance are independent. These are invented teaching fixtures, so a real publisher must replace them with an approved source contract and its own audience research.

Diagnostic case 127: Open-source license notice × Blind paraphrase

Reader-facing test. For a developer reading redistribution guidance, begin with a download page with a license link and do not disclose the intended interpretation. Withhold the source original and ask the reader, in their own language, to explain the message as if describing it to someone who has not seen the screen. The working question is whether the person can distinguish “price, permission and compliance are independent” from the tempting claim that free availability means no obligations. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Listen for the causal story: what triggers the action, what stays true afterward and which actor controls the change. The expected observation is that the person can distinguish free download from authorized redistribution. If instead the answer implies that free availability means no obligations, log the exact words and where the reader looked. The potential harm is unintended redistribution or omission of required notices. Do not treat a fluent reformulation as proof of meaning; ordinary ‘free’ language may mask a contractual distinction. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to surface conditions at the point a user intends to redistribute. A clean paraphrase should preserve the decision rule rather than repeat isolated keywords. A polished but incorrect summary is more informative than a low-confidence guess because it shows the exact rule the reader has constructed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 128: Open-source license notice × Task simulation

Reader-facing test. For a developer reading redistribution guidance, begin with a download page with a license link and do not disclose the intended interpretation. Give a fictional scenario with a choice between acting now, waiting for confirmation and searching for another condition; allow the reader to use only the visible translated material. The working question is whether the person can distinguish “price, permission and compliance are independent” from the tempting claim that free availability means no obligations. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Capture the chosen action, the text used to justify it and whether a wrong branch seems attractive. The expected observation is that the person can distinguish free download from authorized redistribution. If instead the answer implies that free availability means no obligations, log the exact words and where the reader looked. The potential harm is unintended redistribution or omission of required notices. Do not treat a fluent reformulation as proof of meaning; ordinary ‘free’ language may mask a contractual distinction. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to surface conditions at the point a user intends to redistribute. Do not score the final click alone. Someone may guess correctly from placement while holding the wrong belief. A complete result includes the reason, expected consequence and what additional information the person believes is needed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 129: Open-source license notice × Think-aloud trace

Reader-facing test. For a developer reading redistribution guidance, begin with a download page with a license link and do not disclose the intended interpretation. Ask the reader to narrate thoughts while moving from the notice to the relevant action, but do not correct terminology or explain the organization’s policy. The working question is whether the person can distinguish “price, permission and compliance are independent” from the tempting claim that free availability means no obligations. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Mark the moment a term is interpreted, the moment confidence changes and the point at which layout or icons redirect attention. The expected observation is that the person can distinguish free download from authorized redistribution. If instead the answer implies that free availability means no obligations, log the exact words and where the reader looked. The potential harm is unintended redistribution or omission of required notices. Do not treat a fluent reformulation as proof of meaning; ordinary ‘free’ language may mask a contractual distinction. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to surface conditions at the point a user intends to redistribute. This method can identify an interface cue rather than a translation error. The very act of thinking aloud may change behavior, so report it as qualitative diagnostic evidence rather than a naturalistic time-to-completion measurement. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 130: Open-source license notice × Contrastive boundary pair

Reader-facing test. For a developer reading redistribution guidance, begin with a download page with a license link and do not disclose the intended interpretation. Present two nearly identical fictional cases that differ only in the decisive condition; ask whether the outcome changes and why. The working question is whether the person can distinguish “price, permission and compliance are independent” from the tempting claim that free availability means no obligations. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Look for the reader’s treatment of a status, exception or timing boundary that the source intended to preserve. The expected observation is that the person can distinguish free download from authorized redistribution. If instead the answer implies that free availability means no obligations, log the exact words and where the reader looked. The potential harm is unintended redistribution or omission of required notices. Do not treat a fluent reformulation as proof of meaning; ordinary ‘free’ language may mask a contractual distinction. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to surface conditions at the point a user intends to redistribute. A reader who gives the same answer to both cases may have lost the qualifier. However, if the source rule itself is ambiguous, pause the test and clarify it with the content owner. Do not force respondents to guess at a rule the author has not specified. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 131: Open-source license notice × Actor and responsibility map

Reader-facing test. For a developer reading redistribution guidance, begin with a download page with a license link and do not disclose the intended interpretation. Ask the reader to identify who submits, who decides, who can revise the decision and who is affected; let them point to the translated words supporting each answer. The working question is whether the person can distinguish “price, permission and compliance are independent” from the tempting claim that free availability means no obligations. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare the named actors with the interpretation contract and note any responsibility the reader invents. The expected observation is that the person can distinguish free download from authorized redistribution. If instead the answer implies that free availability means no obligations, log the exact words and where the reader looked. The potential harm is unintended redistribution or omission of required notices. Do not treat a fluent reformulation as proof of meaning; ordinary ‘free’ language may mask a contractual distinction. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to surface conditions at the point a user intends to redistribute. Passive voice is not automatically wrong, and some languages omit subjects naturally. The test is whether the actual actor is recoverable. If the reader supplies an unintended actor, identify whether nearby labels, prior notifications or grammatical choices created that inference. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 132: Open-source license notice × Counterexample challenge

Reader-facing test. For a developer reading redistribution guidance, begin with a download page with a license link and do not disclose the intended interpretation. Invite the reader to invent an example in which the notice would not apply, or to say what evidence would make the current instruction no longer valid. The working question is whether the person can distinguish “price, permission and compliance are independent” from the tempting claim that free availability means no obligations. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Inspect whether the reader recognizes limits, exceptions, unknown information and provisional states. The expected observation is that the person can distinguish free download from authorized redistribution. If instead the answer implies that free availability means no obligations, log the exact words and where the reader looked. The potential harm is unintended redistribution or omission of required notices. Do not treat a fluent reformulation as proof of meaning; ordinary ‘free’ language may mask a contractual distinction. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to surface conditions at the point a user intends to redistribute. An inability to imagine a counterexample may be a testing skill issue, so offer a neutral hypothetical rather than marking immediate failure. The key evidence is the reader’s understanding of the boundary after an ordinary scenario is supplied without revealing the answer. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 133: Open-source license notice × Delayed teach-back

Reader-facing test. For a developer reading redistribution guidance, begin with a download page with a license link and do not disclose the intended interpretation. After an unrelated short filler activity, ask the reader to explain the next step to a fictional colleague who cannot view the translated message. The working question is whether the person can distinguish “price, permission and compliance are independent” from the tempting claim that free availability means no obligations. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Note which conditions survive memory, which qualifiers disappear and whether confidence rises as detail is lost. The expected observation is that the person can distinguish free download from authorized redistribution. If instead the answer implies that free availability means no obligations, log the exact words and where the reader looked. The potential harm is unintended redistribution or omission of required notices. Do not treat a fluent reformulation as proof of meaning; ordinary ‘free’ language may mask a contractual distinction. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to surface conditions at the point a user intends to redistribute. Teach-back tests retained practical meaning, not memorization of exact wording. Compare it with first-exposure paraphrase. If the reader initially understood but later drops a critical exception, consider information placement, a reminder, a concise summary or a clearer call to action. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 134: Open-source license notice × Alternative wording comparison

Reader-facing test. For a developer reading redistribution guidance, begin with a download page with a license link and do not disclose the intended interpretation. Show two approved candidate renderings in a balanced order across participants and ask what each would lead them to do; avoid asking which sounds more elegant. The working question is whether the person can distinguish “price, permission and compliance are independent” from the tempting claim that free availability means no obligations. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Record whether the alternatives produce different real-world predictions despite similar fluency judgments. The expected observation is that the person can distinguish free download from authorized redistribution. If instead the answer implies that free availability means no obligations, log the exact words and where the reader looked. The potential harm is unintended redistribution or omission of required notices. Do not treat a fluent reformulation as proof of meaning; ordinary ‘free’ language may mask a contractual distinction. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to surface conditions at the point a user intends to redistribute. Do not optimize language by majority preference alone: a popular label can still induce a serious wrong decision. Preference is a supporting measure. Select wording that better preserves the rule, is natural for the intended audience and survives a subsequent task-based retest. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 135: Open-source license notice × Release-readiness retest

Reader-facing test. For a developer reading redistribution guidance, begin with a download page with a license link and do not disclose the intended interpretation. Present the repaired artifact with the same surrounding interface but an independent fictional case; recruit someone not coached by the earlier test where feasible. The working question is whether the person can distinguish “price, permission and compliance are independent” from the tempting claim that free availability means no obligations. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare unprompted decisions, explanations and remaining errors rather than merely tallying favorable reactions. The expected observation is that the person can distinguish free download from authorized redistribution. If instead the answer implies that free availability means no obligations, log the exact words and where the reader looked. The potential harm is unintended redistribution or omission of required notices. Do not treat a fluent reformulation as proof of meaning; ordinary ‘free’ language may mask a contractual distinction. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to surface conditions at the point a user intends to redistribute. A corrected phrase is not validated by the editor’s confidence. Preserve before-and-after versions, source approval, exact prompts, reader evidence and unresolved risks. Report that the tested failure diminished under these conditions, not that the translation is universally impossible to misunderstand. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Domain studio 16: Privacy opt-out interface

A a reader disabling optional analytics encounters a preference center with several purposes. The intended rule is: Turning off optional measurement does not necessarily disable the account functions that rely on strictly required processing described separately. The risky alternative is opt-out means all processing stops. This matters because an inaccurate understanding of what continues. Here the semantic hinge is that optional and necessary categories must remain distinct. These are invented teaching fixtures, so a real publisher must replace them with an approved source contract and its own audience research.

Diagnostic case 136: Privacy opt-out interface × Blind paraphrase

Reader-facing test. For a reader disabling optional analytics, begin with a preference center with several purposes and do not disclose the intended interpretation. Withhold the source original and ask the reader, in their own language, to explain the message as if describing it to someone who has not seen the screen. The working question is whether the person can distinguish “optional and necessary categories must remain distinct” from the tempting claim that opt-out means all processing stops. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Listen for the causal story: what triggers the action, what stays true afterward and which actor controls the change. The expected observation is that the person names the processing purpose that stops and a different purpose that remains. If instead the answer implies that opt-out means all processing stops, log the exact words and where the reader looked. The potential harm is an inaccurate understanding of what continues. Do not treat a fluent reformulation as proof of meaning; one all-purpose privacy label cannot express several data flows. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to separate each purpose, choice and consequence without implying universal protection. A clean paraphrase should preserve the decision rule rather than repeat isolated keywords. A polished but incorrect summary is more informative than a low-confidence guess because it shows the exact rule the reader has constructed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 137: Privacy opt-out interface × Task simulation

Reader-facing test. For a reader disabling optional analytics, begin with a preference center with several purposes and do not disclose the intended interpretation. Give a fictional scenario with a choice between acting now, waiting for confirmation and searching for another condition; allow the reader to use only the visible translated material. The working question is whether the person can distinguish “optional and necessary categories must remain distinct” from the tempting claim that opt-out means all processing stops. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Capture the chosen action, the text used to justify it and whether a wrong branch seems attractive. The expected observation is that the person names the processing purpose that stops and a different purpose that remains. If instead the answer implies that opt-out means all processing stops, log the exact words and where the reader looked. The potential harm is an inaccurate understanding of what continues. Do not treat a fluent reformulation as proof of meaning; one all-purpose privacy label cannot express several data flows. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to separate each purpose, choice and consequence without implying universal protection. Do not score the final click alone. Someone may guess correctly from placement while holding the wrong belief. A complete result includes the reason, expected consequence and what additional information the person believes is needed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 138: Privacy opt-out interface × Think-aloud trace

Reader-facing test. For a reader disabling optional analytics, begin with a preference center with several purposes and do not disclose the intended interpretation. Ask the reader to narrate thoughts while moving from the notice to the relevant action, but do not correct terminology or explain the organization’s policy. The working question is whether the person can distinguish “optional and necessary categories must remain distinct” from the tempting claim that opt-out means all processing stops. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Mark the moment a term is interpreted, the moment confidence changes and the point at which layout or icons redirect attention. The expected observation is that the person names the processing purpose that stops and a different purpose that remains. If instead the answer implies that opt-out means all processing stops, log the exact words and where the reader looked. The potential harm is an inaccurate understanding of what continues. Do not treat a fluent reformulation as proof of meaning; one all-purpose privacy label cannot express several data flows. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to separate each purpose, choice and consequence without implying universal protection. This method can identify an interface cue rather than a translation error. The very act of thinking aloud may change behavior, so report it as qualitative diagnostic evidence rather than a naturalistic time-to-completion measurement. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 139: Privacy opt-out interface × Contrastive boundary pair

Reader-facing test. For a reader disabling optional analytics, begin with a preference center with several purposes and do not disclose the intended interpretation. Present two nearly identical fictional cases that differ only in the decisive condition; ask whether the outcome changes and why. The working question is whether the person can distinguish “optional and necessary categories must remain distinct” from the tempting claim that opt-out means all processing stops. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Look for the reader’s treatment of a status, exception or timing boundary that the source intended to preserve. The expected observation is that the person names the processing purpose that stops and a different purpose that remains. If instead the answer implies that opt-out means all processing stops, log the exact words and where the reader looked. The potential harm is an inaccurate understanding of what continues. Do not treat a fluent reformulation as proof of meaning; one all-purpose privacy label cannot express several data flows. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to separate each purpose, choice and consequence without implying universal protection. A reader who gives the same answer to both cases may have lost the qualifier. However, if the source rule itself is ambiguous, pause the test and clarify it with the content owner. Do not force respondents to guess at a rule the author has not specified. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 140: Privacy opt-out interface × Actor and responsibility map

Reader-facing test. For a reader disabling optional analytics, begin with a preference center with several purposes and do not disclose the intended interpretation. Ask the reader to identify who submits, who decides, who can revise the decision and who is affected; let them point to the translated words supporting each answer. The working question is whether the person can distinguish “optional and necessary categories must remain distinct” from the tempting claim that opt-out means all processing stops. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare the named actors with the interpretation contract and note any responsibility the reader invents. The expected observation is that the person names the processing purpose that stops and a different purpose that remains. If instead the answer implies that opt-out means all processing stops, log the exact words and where the reader looked. The potential harm is an inaccurate understanding of what continues. Do not treat a fluent reformulation as proof of meaning; one all-purpose privacy label cannot express several data flows. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to separate each purpose, choice and consequence without implying universal protection. Passive voice is not automatically wrong, and some languages omit subjects naturally. The test is whether the actual actor is recoverable. If the reader supplies an unintended actor, identify whether nearby labels, prior notifications or grammatical choices created that inference. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 141: Privacy opt-out interface × Counterexample challenge

Reader-facing test. For a reader disabling optional analytics, begin with a preference center with several purposes and do not disclose the intended interpretation. Invite the reader to invent an example in which the notice would not apply, or to say what evidence would make the current instruction no longer valid. The working question is whether the person can distinguish “optional and necessary categories must remain distinct” from the tempting claim that opt-out means all processing stops. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Inspect whether the reader recognizes limits, exceptions, unknown information and provisional states. The expected observation is that the person names the processing purpose that stops and a different purpose that remains. If instead the answer implies that opt-out means all processing stops, log the exact words and where the reader looked. The potential harm is an inaccurate understanding of what continues. Do not treat a fluent reformulation as proof of meaning; one all-purpose privacy label cannot express several data flows. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to separate each purpose, choice and consequence without implying universal protection. An inability to imagine a counterexample may be a testing skill issue, so offer a neutral hypothetical rather than marking immediate failure. The key evidence is the reader’s understanding of the boundary after an ordinary scenario is supplied without revealing the answer. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 142: Privacy opt-out interface × Delayed teach-back

Reader-facing test. For a reader disabling optional analytics, begin with a preference center with several purposes and do not disclose the intended interpretation. After an unrelated short filler activity, ask the reader to explain the next step to a fictional colleague who cannot view the translated message. The working question is whether the person can distinguish “optional and necessary categories must remain distinct” from the tempting claim that opt-out means all processing stops. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Note which conditions survive memory, which qualifiers disappear and whether confidence rises as detail is lost. The expected observation is that the person names the processing purpose that stops and a different purpose that remains. If instead the answer implies that opt-out means all processing stops, log the exact words and where the reader looked. The potential harm is an inaccurate understanding of what continues. Do not treat a fluent reformulation as proof of meaning; one all-purpose privacy label cannot express several data flows. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to separate each purpose, choice and consequence without implying universal protection. Teach-back tests retained practical meaning, not memorization of exact wording. Compare it with first-exposure paraphrase. If the reader initially understood but later drops a critical exception, consider information placement, a reminder, a concise summary or a clearer call to action. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 143: Privacy opt-out interface × Alternative wording comparison

Reader-facing test. For a reader disabling optional analytics, begin with a preference center with several purposes and do not disclose the intended interpretation. Show two approved candidate renderings in a balanced order across participants and ask what each would lead them to do; avoid asking which sounds more elegant. The working question is whether the person can distinguish “optional and necessary categories must remain distinct” from the tempting claim that opt-out means all processing stops. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Record whether the alternatives produce different real-world predictions despite similar fluency judgments. The expected observation is that the person names the processing purpose that stops and a different purpose that remains. If instead the answer implies that opt-out means all processing stops, log the exact words and where the reader looked. The potential harm is an inaccurate understanding of what continues. Do not treat a fluent reformulation as proof of meaning; one all-purpose privacy label cannot express several data flows. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to separate each purpose, choice and consequence without implying universal protection. Do not optimize language by majority preference alone: a popular label can still induce a serious wrong decision. Preference is a supporting measure. Select wording that better preserves the rule, is natural for the intended audience and survives a subsequent task-based retest. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 144: Privacy opt-out interface × Release-readiness retest

Reader-facing test. For a reader disabling optional analytics, begin with a preference center with several purposes and do not disclose the intended interpretation. Present the repaired artifact with the same surrounding interface but an independent fictional case; recruit someone not coached by the earlier test where feasible. The working question is whether the person can distinguish “optional and necessary categories must remain distinct” from the tempting claim that opt-out means all processing stops. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare unprompted decisions, explanations and remaining errors rather than merely tallying favorable reactions. The expected observation is that the person names the processing purpose that stops and a different purpose that remains. If instead the answer implies that opt-out means all processing stops, log the exact words and where the reader looked. The potential harm is an inaccurate understanding of what continues. Do not treat a fluent reformulation as proof of meaning; one all-purpose privacy label cannot express several data flows. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to separate each purpose, choice and consequence without implying universal protection. A corrected phrase is not validated by the editor’s confidence. Preserve before-and-after versions, source approval, exact prompts, reader evidence and unresolved risks. Report that the tested failure diminished under these conditions, not that the translation is universally impossible to misunderstand. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Domain studio 17: Card dispute timeline

A a cardholder looking at a disputed transaction encounters a dispute status panel with transaction amount. The intended rule is: Submitting a dispute opens a review, while any temporary credit is provisional until a final determination. The risky alternative is provisional credit is a final refund. This matters because false expectations about available funds. Here the semantic hinge is that interim balance treatment is not a final adjudication. These are invented teaching fixtures, so a real publisher must replace them with an approved source contract and its own audience research.

Diagnostic case 145: Card dispute timeline × Blind paraphrase

Reader-facing test. For a cardholder looking at a disputed transaction, begin with a dispute status panel with transaction amount and do not disclose the intended interpretation. Withhold the source original and ask the reader, in their own language, to explain the message as if describing it to someone who has not seen the screen. The working question is whether the person can distinguish “interim balance treatment is not a final adjudication” from the tempting claim that provisional credit is a final refund. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Listen for the causal story: what triggers the action, what stays true afterward and which actor controls the change. The expected observation is that the person can distinguish investigation from resolution. If instead the answer implies that provisional credit is a final refund, log the exact words and where the reader looked. The potential harm is false expectations about available funds. Do not treat a fluent reformulation as proof of meaning; a visible balance change may not settle the underlying claim. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to name the provisional nature and the next decision point. A clean paraphrase should preserve the decision rule rather than repeat isolated keywords. A polished but incorrect summary is more informative than a low-confidence guess because it shows the exact rule the reader has constructed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 146: Card dispute timeline × Task simulation

Reader-facing test. For a cardholder looking at a disputed transaction, begin with a dispute status panel with transaction amount and do not disclose the intended interpretation. Give a fictional scenario with a choice between acting now, waiting for confirmation and searching for another condition; allow the reader to use only the visible translated material. The working question is whether the person can distinguish “interim balance treatment is not a final adjudication” from the tempting claim that provisional credit is a final refund. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Capture the chosen action, the text used to justify it and whether a wrong branch seems attractive. The expected observation is that the person can distinguish investigation from resolution. If instead the answer implies that provisional credit is a final refund, log the exact words and where the reader looked. The potential harm is false expectations about available funds. Do not treat a fluent reformulation as proof of meaning; a visible balance change may not settle the underlying claim. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to name the provisional nature and the next decision point. Do not score the final click alone. Someone may guess correctly from placement while holding the wrong belief. A complete result includes the reason, expected consequence and what additional information the person believes is needed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 147: Card dispute timeline × Think-aloud trace

Reader-facing test. For a cardholder looking at a disputed transaction, begin with a dispute status panel with transaction amount and do not disclose the intended interpretation. Ask the reader to narrate thoughts while moving from the notice to the relevant action, but do not correct terminology or explain the organization’s policy. The working question is whether the person can distinguish “interim balance treatment is not a final adjudication” from the tempting claim that provisional credit is a final refund. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Mark the moment a term is interpreted, the moment confidence changes and the point at which layout or icons redirect attention. The expected observation is that the person can distinguish investigation from resolution. If instead the answer implies that provisional credit is a final refund, log the exact words and where the reader looked. The potential harm is false expectations about available funds. Do not treat a fluent reformulation as proof of meaning; a visible balance change may not settle the underlying claim. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to name the provisional nature and the next decision point. This method can identify an interface cue rather than a translation error. The very act of thinking aloud may change behavior, so report it as qualitative diagnostic evidence rather than a naturalistic time-to-completion measurement. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 148: Card dispute timeline × Contrastive boundary pair

Reader-facing test. For a cardholder looking at a disputed transaction, begin with a dispute status panel with transaction amount and do not disclose the intended interpretation. Present two nearly identical fictional cases that differ only in the decisive condition; ask whether the outcome changes and why. The working question is whether the person can distinguish “interim balance treatment is not a final adjudication” from the tempting claim that provisional credit is a final refund. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Look for the reader’s treatment of a status, exception or timing boundary that the source intended to preserve. The expected observation is that the person can distinguish investigation from resolution. If instead the answer implies that provisional credit is a final refund, log the exact words and where the reader looked. The potential harm is false expectations about available funds. Do not treat a fluent reformulation as proof of meaning; a visible balance change may not settle the underlying claim. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to name the provisional nature and the next decision point. A reader who gives the same answer to both cases may have lost the qualifier. However, if the source rule itself is ambiguous, pause the test and clarify it with the content owner. Do not force respondents to guess at a rule the author has not specified. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 149: Card dispute timeline × Actor and responsibility map

Reader-facing test. For a cardholder looking at a disputed transaction, begin with a dispute status panel with transaction amount and do not disclose the intended interpretation. Ask the reader to identify who submits, who decides, who can revise the decision and who is affected; let them point to the translated words supporting each answer. The working question is whether the person can distinguish “interim balance treatment is not a final adjudication” from the tempting claim that provisional credit is a final refund. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare the named actors with the interpretation contract and note any responsibility the reader invents. The expected observation is that the person can distinguish investigation from resolution. If instead the answer implies that provisional credit is a final refund, log the exact words and where the reader looked. The potential harm is false expectations about available funds. Do not treat a fluent reformulation as proof of meaning; a visible balance change may not settle the underlying claim. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to name the provisional nature and the next decision point. Passive voice is not automatically wrong, and some languages omit subjects naturally. The test is whether the actual actor is recoverable. If the reader supplies an unintended actor, identify whether nearby labels, prior notifications or grammatical choices created that inference. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 150: Card dispute timeline × Counterexample challenge

Reader-facing test. For a cardholder looking at a disputed transaction, begin with a dispute status panel with transaction amount and do not disclose the intended interpretation. Invite the reader to invent an example in which the notice would not apply, or to say what evidence would make the current instruction no longer valid. The working question is whether the person can distinguish “interim balance treatment is not a final adjudication” from the tempting claim that provisional credit is a final refund. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Inspect whether the reader recognizes limits, exceptions, unknown information and provisional states. The expected observation is that the person can distinguish investigation from resolution. If instead the answer implies that provisional credit is a final refund, log the exact words and where the reader looked. The potential harm is false expectations about available funds. Do not treat a fluent reformulation as proof of meaning; a visible balance change may not settle the underlying claim. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to name the provisional nature and the next decision point. An inability to imagine a counterexample may be a testing skill issue, so offer a neutral hypothetical rather than marking immediate failure. The key evidence is the reader’s understanding of the boundary after an ordinary scenario is supplied without revealing the answer. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 151: Card dispute timeline × Delayed teach-back

Reader-facing test. For a cardholder looking at a disputed transaction, begin with a dispute status panel with transaction amount and do not disclose the intended interpretation. After an unrelated short filler activity, ask the reader to explain the next step to a fictional colleague who cannot view the translated message. The working question is whether the person can distinguish “interim balance treatment is not a final adjudication” from the tempting claim that provisional credit is a final refund. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Note which conditions survive memory, which qualifiers disappear and whether confidence rises as detail is lost. The expected observation is that the person can distinguish investigation from resolution. If instead the answer implies that provisional credit is a final refund, log the exact words and where the reader looked. The potential harm is false expectations about available funds. Do not treat a fluent reformulation as proof of meaning; a visible balance change may not settle the underlying claim. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to name the provisional nature and the next decision point. Teach-back tests retained practical meaning, not memorization of exact wording. Compare it with first-exposure paraphrase. If the reader initially understood but later drops a critical exception, consider information placement, a reminder, a concise summary or a clearer call to action. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 152: Card dispute timeline × Alternative wording comparison

Reader-facing test. For a cardholder looking at a disputed transaction, begin with a dispute status panel with transaction amount and do not disclose the intended interpretation. Show two approved candidate renderings in a balanced order across participants and ask what each would lead them to do; avoid asking which sounds more elegant. The working question is whether the person can distinguish “interim balance treatment is not a final adjudication” from the tempting claim that provisional credit is a final refund. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Record whether the alternatives produce different real-world predictions despite similar fluency judgments. The expected observation is that the person can distinguish investigation from resolution. If instead the answer implies that provisional credit is a final refund, log the exact words and where the reader looked. The potential harm is false expectations about available funds. Do not treat a fluent reformulation as proof of meaning; a visible balance change may not settle the underlying claim. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to name the provisional nature and the next decision point. Do not optimize language by majority preference alone: a popular label can still induce a serious wrong decision. Preference is a supporting measure. Select wording that better preserves the rule, is natural for the intended audience and survives a subsequent task-based retest. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 153: Card dispute timeline × Release-readiness retest

Reader-facing test. For a cardholder looking at a disputed transaction, begin with a dispute status panel with transaction amount and do not disclose the intended interpretation. Present the repaired artifact with the same surrounding interface but an independent fictional case; recruit someone not coached by the earlier test where feasible. The working question is whether the person can distinguish “interim balance treatment is not a final adjudication” from the tempting claim that provisional credit is a final refund. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare unprompted decisions, explanations and remaining errors rather than merely tallying favorable reactions. The expected observation is that the person can distinguish investigation from resolution. If instead the answer implies that provisional credit is a final refund, log the exact words and where the reader looked. The potential harm is false expectations about available funds. Do not treat a fluent reformulation as proof of meaning; a visible balance change may not settle the underlying claim. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to name the provisional nature and the next decision point. A corrected phrase is not validated by the editor’s confidence. Preserve before-and-after versions, source approval, exact prompts, reader evidence and unresolved risks. Report that the tested failure diminished under these conditions, not that the translation is universally impossible to misunderstand. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Domain studio 18: Food allergen menu notice

A a customer with a dietary restriction reviewing menu information encounters a menu description and a separate allergy-information route. The intended rule is: An ingredient list does not by itself establish that food was prepared without cross-contact; relevant allergy information requires confirmation from the provider. The risky alternative is absence from ingredient list guarantees absence of allergen exposure. This matters because a potentially serious unsafe inference. Here the semantic hinge is that composition and preparation environment are different evidence. These are invented teaching fixtures, so a real publisher must replace them with an approved source contract and its own audience research.

Diagnostic case 154: Food allergen menu notice × Blind paraphrase

Reader-facing test. For a customer with a dietary restriction reviewing menu information, begin with a menu description and a separate allergy-information route and do not disclose the intended interpretation. Withhold the source original and ask the reader, in their own language, to explain the message as if describing it to someone who has not seen the screen. The working question is whether the person can distinguish “composition and preparation environment are different evidence” from the tempting claim that absence from ingredient list guarantees absence of allergen exposure. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Listen for the causal story: what triggers the action, what stays true afterward and which actor controls the change. The expected observation is that the person recognizes that missing ingredient text is not definitive safety proof. If instead the answer implies that absence from ingredient list guarantees absence of allergen exposure, log the exact words and where the reader looked. The potential harm is a potentially serious unsafe inference. Do not treat a fluent reformulation as proof of meaning; negative evidence can be overread as proof of absence. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to avoid certainty that the evidence cannot support and refer to verified provider information. A clean paraphrase should preserve the decision rule rather than repeat isolated keywords. A polished but incorrect summary is more informative than a low-confidence guess because it shows the exact rule the reader has constructed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 155: Food allergen menu notice × Task simulation

Reader-facing test. For a customer with a dietary restriction reviewing menu information, begin with a menu description and a separate allergy-information route and do not disclose the intended interpretation. Give a fictional scenario with a choice between acting now, waiting for confirmation and searching for another condition; allow the reader to use only the visible translated material. The working question is whether the person can distinguish “composition and preparation environment are different evidence” from the tempting claim that absence from ingredient list guarantees absence of allergen exposure. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Capture the chosen action, the text used to justify it and whether a wrong branch seems attractive. The expected observation is that the person recognizes that missing ingredient text is not definitive safety proof. If instead the answer implies that absence from ingredient list guarantees absence of allergen exposure, log the exact words and where the reader looked. The potential harm is a potentially serious unsafe inference. Do not treat a fluent reformulation as proof of meaning; negative evidence can be overread as proof of absence. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to avoid certainty that the evidence cannot support and refer to verified provider information. Do not score the final click alone. Someone may guess correctly from placement while holding the wrong belief. A complete result includes the reason, expected consequence and what additional information the person believes is needed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 156: Food allergen menu notice × Think-aloud trace

Reader-facing test. For a customer with a dietary restriction reviewing menu information, begin with a menu description and a separate allergy-information route and do not disclose the intended interpretation. Ask the reader to narrate thoughts while moving from the notice to the relevant action, but do not correct terminology or explain the organization’s policy. The working question is whether the person can distinguish “composition and preparation environment are different evidence” from the tempting claim that absence from ingredient list guarantees absence of allergen exposure. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Mark the moment a term is interpreted, the moment confidence changes and the point at which layout or icons redirect attention. The expected observation is that the person recognizes that missing ingredient text is not definitive safety proof. If instead the answer implies that absence from ingredient list guarantees absence of allergen exposure, log the exact words and where the reader looked. The potential harm is a potentially serious unsafe inference. Do not treat a fluent reformulation as proof of meaning; negative evidence can be overread as proof of absence. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to avoid certainty that the evidence cannot support and refer to verified provider information. This method can identify an interface cue rather than a translation error. The very act of thinking aloud may change behavior, so report it as qualitative diagnostic evidence rather than a naturalistic time-to-completion measurement. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 157: Food allergen menu notice × Contrastive boundary pair

Reader-facing test. For a customer with a dietary restriction reviewing menu information, begin with a menu description and a separate allergy-information route and do not disclose the intended interpretation. Present two nearly identical fictional cases that differ only in the decisive condition; ask whether the outcome changes and why. The working question is whether the person can distinguish “composition and preparation environment are different evidence” from the tempting claim that absence from ingredient list guarantees absence of allergen exposure. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Look for the reader’s treatment of a status, exception or timing boundary that the source intended to preserve. The expected observation is that the person recognizes that missing ingredient text is not definitive safety proof. If instead the answer implies that absence from ingredient list guarantees absence of allergen exposure, log the exact words and where the reader looked. The potential harm is a potentially serious unsafe inference. Do not treat a fluent reformulation as proof of meaning; negative evidence can be overread as proof of absence. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to avoid certainty that the evidence cannot support and refer to verified provider information. A reader who gives the same answer to both cases may have lost the qualifier. However, if the source rule itself is ambiguous, pause the test and clarify it with the content owner. Do not force respondents to guess at a rule the author has not specified. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 158: Food allergen menu notice × Actor and responsibility map

Reader-facing test. For a customer with a dietary restriction reviewing menu information, begin with a menu description and a separate allergy-information route and do not disclose the intended interpretation. Ask the reader to identify who submits, who decides, who can revise the decision and who is affected; let them point to the translated words supporting each answer. The working question is whether the person can distinguish “composition and preparation environment are different evidence” from the tempting claim that absence from ingredient list guarantees absence of allergen exposure. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare the named actors with the interpretation contract and note any responsibility the reader invents. The expected observation is that the person recognizes that missing ingredient text is not definitive safety proof. If instead the answer implies that absence from ingredient list guarantees absence of allergen exposure, log the exact words and where the reader looked. The potential harm is a potentially serious unsafe inference. Do not treat a fluent reformulation as proof of meaning; negative evidence can be overread as proof of absence. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to avoid certainty that the evidence cannot support and refer to verified provider information. Passive voice is not automatically wrong, and some languages omit subjects naturally. The test is whether the actual actor is recoverable. If the reader supplies an unintended actor, identify whether nearby labels, prior notifications or grammatical choices created that inference. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 159: Food allergen menu notice × Counterexample challenge

Reader-facing test. For a customer with a dietary restriction reviewing menu information, begin with a menu description and a separate allergy-information route and do not disclose the intended interpretation. Invite the reader to invent an example in which the notice would not apply, or to say what evidence would make the current instruction no longer valid. The working question is whether the person can distinguish “composition and preparation environment are different evidence” from the tempting claim that absence from ingredient list guarantees absence of allergen exposure. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Inspect whether the reader recognizes limits, exceptions, unknown information and provisional states. The expected observation is that the person recognizes that missing ingredient text is not definitive safety proof. If instead the answer implies that absence from ingredient list guarantees absence of allergen exposure, log the exact words and where the reader looked. The potential harm is a potentially serious unsafe inference. Do not treat a fluent reformulation as proof of meaning; negative evidence can be overread as proof of absence. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to avoid certainty that the evidence cannot support and refer to verified provider information. An inability to imagine a counterexample may be a testing skill issue, so offer a neutral hypothetical rather than marking immediate failure. The key evidence is the reader’s understanding of the boundary after an ordinary scenario is supplied without revealing the answer. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 160: Food allergen menu notice × Delayed teach-back

Reader-facing test. For a customer with a dietary restriction reviewing menu information, begin with a menu description and a separate allergy-information route and do not disclose the intended interpretation. After an unrelated short filler activity, ask the reader to explain the next step to a fictional colleague who cannot view the translated message. The working question is whether the person can distinguish “composition and preparation environment are different evidence” from the tempting claim that absence from ingredient list guarantees absence of allergen exposure. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Note which conditions survive memory, which qualifiers disappear and whether confidence rises as detail is lost. The expected observation is that the person recognizes that missing ingredient text is not definitive safety proof. If instead the answer implies that absence from ingredient list guarantees absence of allergen exposure, log the exact words and where the reader looked. The potential harm is a potentially serious unsafe inference. Do not treat a fluent reformulation as proof of meaning; negative evidence can be overread as proof of absence. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to avoid certainty that the evidence cannot support and refer to verified provider information. Teach-back tests retained practical meaning, not memorization of exact wording. Compare it with first-exposure paraphrase. If the reader initially understood but later drops a critical exception, consider information placement, a reminder, a concise summary or a clearer call to action. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 161: Food allergen menu notice × Alternative wording comparison

Reader-facing test. For a customer with a dietary restriction reviewing menu information, begin with a menu description and a separate allergy-information route and do not disclose the intended interpretation. Show two approved candidate renderings in a balanced order across participants and ask what each would lead them to do; avoid asking which sounds more elegant. The working question is whether the person can distinguish “composition and preparation environment are different evidence” from the tempting claim that absence from ingredient list guarantees absence of allergen exposure. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Record whether the alternatives produce different real-world predictions despite similar fluency judgments. The expected observation is that the person recognizes that missing ingredient text is not definitive safety proof. If instead the answer implies that absence from ingredient list guarantees absence of allergen exposure, log the exact words and where the reader looked. The potential harm is a potentially serious unsafe inference. Do not treat a fluent reformulation as proof of meaning; negative evidence can be overread as proof of absence. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to avoid certainty that the evidence cannot support and refer to verified provider information. Do not optimize language by majority preference alone: a popular label can still induce a serious wrong decision. Preference is a supporting measure. Select wording that better preserves the rule, is natural for the intended audience and survives a subsequent task-based retest. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 162: Food allergen menu notice × Release-readiness retest

Reader-facing test. For a customer with a dietary restriction reviewing menu information, begin with a menu description and a separate allergy-information route and do not disclose the intended interpretation. Present the repaired artifact with the same surrounding interface but an independent fictional case; recruit someone not coached by the earlier test where feasible. The working question is whether the person can distinguish “composition and preparation environment are different evidence” from the tempting claim that absence from ingredient list guarantees absence of allergen exposure. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare unprompted decisions, explanations and remaining errors rather than merely tallying favorable reactions. The expected observation is that the person recognizes that missing ingredient text is not definitive safety proof. If instead the answer implies that absence from ingredient list guarantees absence of allergen exposure, log the exact words and where the reader looked. The potential harm is a potentially serious unsafe inference. Do not treat a fluent reformulation as proof of meaning; negative evidence can be overread as proof of absence. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to avoid certainty that the evidence cannot support and refer to verified provider information. A corrected phrase is not validated by the editor’s confidence. Preserve before-and-after versions, source approval, exact prompts, reader evidence and unresolved risks. Report that the tested failure diminished under these conditions, not that the translation is universally impossible to misunderstand. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Domain studio 19: Automated moderation appeal

A a creator seeing restricted-content status encounters a restricted badge and appeal form. The intended rule is: A notice records an automated restriction and provides an appeal route; filing the appeal does not automatically restore the content. The risky alternative is appeal submission guarantees immediate reinstatement. This matters because mistaken claims of restored visibility. Here the semantic hinge is that review initiation differs from decision reversal. These are invented teaching fixtures, so a real publisher must replace them with an approved source contract and its own audience research.

Diagnostic case 163: Automated moderation appeal × Blind paraphrase

Reader-facing test. For a creator seeing restricted-content status, begin with a restricted badge and appeal form and do not disclose the intended interpretation. Withhold the source original and ask the reader, in their own language, to explain the message as if describing it to someone who has not seen the screen. The working question is whether the person can distinguish “review initiation differs from decision reversal” from the tempting claim that appeal submission guarantees immediate reinstatement. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Listen for the causal story: what triggers the action, what stays true afterward and which actor controls the change. The expected observation is that the person can predict what filing an appeal does and does not do. If instead the answer implies that appeal submission guarantees immediate reinstatement, log the exact words and where the reader looked. The potential harm is mistaken claims of restored visibility. Do not treat a fluent reformulation as proof of meaning; procedural recourse and outcome are not synonyms. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to display appeal received, under review and resolved as separate states. A clean paraphrase should preserve the decision rule rather than repeat isolated keywords. A polished but incorrect summary is more informative than a low-confidence guess because it shows the exact rule the reader has constructed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 164: Automated moderation appeal × Task simulation

Reader-facing test. For a creator seeing restricted-content status, begin with a restricted badge and appeal form and do not disclose the intended interpretation. Give a fictional scenario with a choice between acting now, waiting for confirmation and searching for another condition; allow the reader to use only the visible translated material. The working question is whether the person can distinguish “review initiation differs from decision reversal” from the tempting claim that appeal submission guarantees immediate reinstatement. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Capture the chosen action, the text used to justify it and whether a wrong branch seems attractive. The expected observation is that the person can predict what filing an appeal does and does not do. If instead the answer implies that appeal submission guarantees immediate reinstatement, log the exact words and where the reader looked. The potential harm is mistaken claims of restored visibility. Do not treat a fluent reformulation as proof of meaning; procedural recourse and outcome are not synonyms. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to display appeal received, under review and resolved as separate states. Do not score the final click alone. Someone may guess correctly from placement while holding the wrong belief. A complete result includes the reason, expected consequence and what additional information the person believes is needed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 165: Automated moderation appeal × Think-aloud trace

Reader-facing test. For a creator seeing restricted-content status, begin with a restricted badge and appeal form and do not disclose the intended interpretation. Ask the reader to narrate thoughts while moving from the notice to the relevant action, but do not correct terminology or explain the organization’s policy. The working question is whether the person can distinguish “review initiation differs from decision reversal” from the tempting claim that appeal submission guarantees immediate reinstatement. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Mark the moment a term is interpreted, the moment confidence changes and the point at which layout or icons redirect attention. The expected observation is that the person can predict what filing an appeal does and does not do. If instead the answer implies that appeal submission guarantees immediate reinstatement, log the exact words and where the reader looked. The potential harm is mistaken claims of restored visibility. Do not treat a fluent reformulation as proof of meaning; procedural recourse and outcome are not synonyms. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to display appeal received, under review and resolved as separate states. This method can identify an interface cue rather than a translation error. The very act of thinking aloud may change behavior, so report it as qualitative diagnostic evidence rather than a naturalistic time-to-completion measurement. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 166: Automated moderation appeal × Contrastive boundary pair

Reader-facing test. For a creator seeing restricted-content status, begin with a restricted badge and appeal form and do not disclose the intended interpretation. Present two nearly identical fictional cases that differ only in the decisive condition; ask whether the outcome changes and why. The working question is whether the person can distinguish “review initiation differs from decision reversal” from the tempting claim that appeal submission guarantees immediate reinstatement. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Look for the reader’s treatment of a status, exception or timing boundary that the source intended to preserve. The expected observation is that the person can predict what filing an appeal does and does not do. If instead the answer implies that appeal submission guarantees immediate reinstatement, log the exact words and where the reader looked. The potential harm is mistaken claims of restored visibility. Do not treat a fluent reformulation as proof of meaning; procedural recourse and outcome are not synonyms. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to display appeal received, under review and resolved as separate states. A reader who gives the same answer to both cases may have lost the qualifier. However, if the source rule itself is ambiguous, pause the test and clarify it with the content owner. Do not force respondents to guess at a rule the author has not specified. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 167: Automated moderation appeal × Actor and responsibility map

Reader-facing test. For a creator seeing restricted-content status, begin with a restricted badge and appeal form and do not disclose the intended interpretation. Ask the reader to identify who submits, who decides, who can revise the decision and who is affected; let them point to the translated words supporting each answer. The working question is whether the person can distinguish “review initiation differs from decision reversal” from the tempting claim that appeal submission guarantees immediate reinstatement. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare the named actors with the interpretation contract and note any responsibility the reader invents. The expected observation is that the person can predict what filing an appeal does and does not do. If instead the answer implies that appeal submission guarantees immediate reinstatement, log the exact words and where the reader looked. The potential harm is mistaken claims of restored visibility. Do not treat a fluent reformulation as proof of meaning; procedural recourse and outcome are not synonyms. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to display appeal received, under review and resolved as separate states. Passive voice is not automatically wrong, and some languages omit subjects naturally. The test is whether the actual actor is recoverable. If the reader supplies an unintended actor, identify whether nearby labels, prior notifications or grammatical choices created that inference. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 168: Automated moderation appeal × Counterexample challenge

Reader-facing test. For a creator seeing restricted-content status, begin with a restricted badge and appeal form and do not disclose the intended interpretation. Invite the reader to invent an example in which the notice would not apply, or to say what evidence would make the current instruction no longer valid. The working question is whether the person can distinguish “review initiation differs from decision reversal” from the tempting claim that appeal submission guarantees immediate reinstatement. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Inspect whether the reader recognizes limits, exceptions, unknown information and provisional states. The expected observation is that the person can predict what filing an appeal does and does not do. If instead the answer implies that appeal submission guarantees immediate reinstatement, log the exact words and where the reader looked. The potential harm is mistaken claims of restored visibility. Do not treat a fluent reformulation as proof of meaning; procedural recourse and outcome are not synonyms. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to display appeal received, under review and resolved as separate states. An inability to imagine a counterexample may be a testing skill issue, so offer a neutral hypothetical rather than marking immediate failure. The key evidence is the reader’s understanding of the boundary after an ordinary scenario is supplied without revealing the answer. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 169: Automated moderation appeal × Delayed teach-back

Reader-facing test. For a creator seeing restricted-content status, begin with a restricted badge and appeal form and do not disclose the intended interpretation. After an unrelated short filler activity, ask the reader to explain the next step to a fictional colleague who cannot view the translated message. The working question is whether the person can distinguish “review initiation differs from decision reversal” from the tempting claim that appeal submission guarantees immediate reinstatement. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Note which conditions survive memory, which qualifiers disappear and whether confidence rises as detail is lost. The expected observation is that the person can predict what filing an appeal does and does not do. If instead the answer implies that appeal submission guarantees immediate reinstatement, log the exact words and where the reader looked. The potential harm is mistaken claims of restored visibility. Do not treat a fluent reformulation as proof of meaning; procedural recourse and outcome are not synonyms. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to display appeal received, under review and resolved as separate states. Teach-back tests retained practical meaning, not memorization of exact wording. Compare it with first-exposure paraphrase. If the reader initially understood but later drops a critical exception, consider information placement, a reminder, a concise summary or a clearer call to action. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 170: Automated moderation appeal × Alternative wording comparison

Reader-facing test. For a creator seeing restricted-content status, begin with a restricted badge and appeal form and do not disclose the intended interpretation. Show two approved candidate renderings in a balanced order across participants and ask what each would lead them to do; avoid asking which sounds more elegant. The working question is whether the person can distinguish “review initiation differs from decision reversal” from the tempting claim that appeal submission guarantees immediate reinstatement. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Record whether the alternatives produce different real-world predictions despite similar fluency judgments. The expected observation is that the person can predict what filing an appeal does and does not do. If instead the answer implies that appeal submission guarantees immediate reinstatement, log the exact words and where the reader looked. The potential harm is mistaken claims of restored visibility. Do not treat a fluent reformulation as proof of meaning; procedural recourse and outcome are not synonyms. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to display appeal received, under review and resolved as separate states. Do not optimize language by majority preference alone: a popular label can still induce a serious wrong decision. Preference is a supporting measure. Select wording that better preserves the rule, is natural for the intended audience and survives a subsequent task-based retest. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 171: Automated moderation appeal × Release-readiness retest

Reader-facing test. For a creator seeing restricted-content status, begin with a restricted badge and appeal form and do not disclose the intended interpretation. Present the repaired artifact with the same surrounding interface but an independent fictional case; recruit someone not coached by the earlier test where feasible. The working question is whether the person can distinguish “review initiation differs from decision reversal” from the tempting claim that appeal submission guarantees immediate reinstatement. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare unprompted decisions, explanations and remaining errors rather than merely tallying favorable reactions. The expected observation is that the person can predict what filing an appeal does and does not do. If instead the answer implies that appeal submission guarantees immediate reinstatement, log the exact words and where the reader looked. The potential harm is mistaken claims of restored visibility. Do not treat a fluent reformulation as proof of meaning; procedural recourse and outcome are not synonyms. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to display appeal received, under review and resolved as separate states. A corrected phrase is not validated by the editor’s confidence. Preserve before-and-after versions, source approval, exact prompts, reader evidence and unresolved risks. Report that the tested failure diminished under these conditions, not that the translation is universally impossible to misunderstand. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Domain studio 20: Research participation consent

A a volunteer reading a study information sheet encounters a consent form next to the ordinary sign-in process. The intended rule is: The person may decline participation without losing access to ordinary unrelated services, and consent can be withdrawn subject to the disclosed data handling terms. The risky alternative is declining means losing all access or withdrawal erases every existing record. This matters because coerced participation or incorrect data expectations. Here the semantic hinge is that voluntariness and past data handling are independent dimensions. These are invented teaching fixtures, so a real publisher must replace them with an approved source contract and its own audience research.

Diagnostic case 172: Research participation consent × Blind paraphrase

Reader-facing test. For a volunteer reading a study information sheet, begin with a consent form next to the ordinary sign-in process and do not disclose the intended interpretation. Withhold the source original and ask the reader, in their own language, to explain the message as if describing it to someone who has not seen the screen. The working question is whether the person can distinguish “voluntariness and past data handling are independent dimensions” from the tempting claim that declining means losing all access or withdrawal erases every existing record. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Listen for the causal story: what triggers the action, what stays true afterward and which actor controls the change. The expected observation is that the person can explain both the freedom to decline and the limits of withdrawal. If instead the answer implies that declining means losing all access or withdrawal erases every existing record, log the exact words and where the reader looked. The potential harm is coerced participation or incorrect data expectations. Do not treat a fluent reformulation as proof of meaning; a single polite sentence may conceal two separate rights. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to state the independent services and the documented withdrawal consequences. A clean paraphrase should preserve the decision rule rather than repeat isolated keywords. A polished but incorrect summary is more informative than a low-confidence guess because it shows the exact rule the reader has constructed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 173: Research participation consent × Task simulation

Reader-facing test. For a volunteer reading a study information sheet, begin with a consent form next to the ordinary sign-in process and do not disclose the intended interpretation. Give a fictional scenario with a choice between acting now, waiting for confirmation and searching for another condition; allow the reader to use only the visible translated material. The working question is whether the person can distinguish “voluntariness and past data handling are independent dimensions” from the tempting claim that declining means losing all access or withdrawal erases every existing record. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Capture the chosen action, the text used to justify it and whether a wrong branch seems attractive. The expected observation is that the person can explain both the freedom to decline and the limits of withdrawal. If instead the answer implies that declining means losing all access or withdrawal erases every existing record, log the exact words and where the reader looked. The potential harm is coerced participation or incorrect data expectations. Do not treat a fluent reformulation as proof of meaning; a single polite sentence may conceal two separate rights. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to state the independent services and the documented withdrawal consequences. Do not score the final click alone. Someone may guess correctly from placement while holding the wrong belief. A complete result includes the reason, expected consequence and what additional information the person believes is needed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 174: Research participation consent × Think-aloud trace

Reader-facing test. For a volunteer reading a study information sheet, begin with a consent form next to the ordinary sign-in process and do not disclose the intended interpretation. Ask the reader to narrate thoughts while moving from the notice to the relevant action, but do not correct terminology or explain the organization’s policy. The working question is whether the person can distinguish “voluntariness and past data handling are independent dimensions” from the tempting claim that declining means losing all access or withdrawal erases every existing record. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Mark the moment a term is interpreted, the moment confidence changes and the point at which layout or icons redirect attention. The expected observation is that the person can explain both the freedom to decline and the limits of withdrawal. If instead the answer implies that declining means losing all access or withdrawal erases every existing record, log the exact words and where the reader looked. The potential harm is coerced participation or incorrect data expectations. Do not treat a fluent reformulation as proof of meaning; a single polite sentence may conceal two separate rights. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to state the independent services and the documented withdrawal consequences. This method can identify an interface cue rather than a translation error. The very act of thinking aloud may change behavior, so report it as qualitative diagnostic evidence rather than a naturalistic time-to-completion measurement. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 175: Research participation consent × Contrastive boundary pair

Reader-facing test. For a volunteer reading a study information sheet, begin with a consent form next to the ordinary sign-in process and do not disclose the intended interpretation. Present two nearly identical fictional cases that differ only in the decisive condition; ask whether the outcome changes and why. The working question is whether the person can distinguish “voluntariness and past data handling are independent dimensions” from the tempting claim that declining means losing all access or withdrawal erases every existing record. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Look for the reader’s treatment of a status, exception or timing boundary that the source intended to preserve. The expected observation is that the person can explain both the freedom to decline and the limits of withdrawal. If instead the answer implies that declining means losing all access or withdrawal erases every existing record, log the exact words and where the reader looked. The potential harm is coerced participation or incorrect data expectations. Do not treat a fluent reformulation as proof of meaning; a single polite sentence may conceal two separate rights. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to state the independent services and the documented withdrawal consequences. A reader who gives the same answer to both cases may have lost the qualifier. However, if the source rule itself is ambiguous, pause the test and clarify it with the content owner. Do not force respondents to guess at a rule the author has not specified. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 176: Research participation consent × Actor and responsibility map

Reader-facing test. For a volunteer reading a study information sheet, begin with a consent form next to the ordinary sign-in process and do not disclose the intended interpretation. Ask the reader to identify who submits, who decides, who can revise the decision and who is affected; let them point to the translated words supporting each answer. The working question is whether the person can distinguish “voluntariness and past data handling are independent dimensions” from the tempting claim that declining means losing all access or withdrawal erases every existing record. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare the named actors with the interpretation contract and note any responsibility the reader invents. The expected observation is that the person can explain both the freedom to decline and the limits of withdrawal. If instead the answer implies that declining means losing all access or withdrawal erases every existing record, log the exact words and where the reader looked. The potential harm is coerced participation or incorrect data expectations. Do not treat a fluent reformulation as proof of meaning; a single polite sentence may conceal two separate rights. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to state the independent services and the documented withdrawal consequences. Passive voice is not automatically wrong, and some languages omit subjects naturally. The test is whether the actual actor is recoverable. If the reader supplies an unintended actor, identify whether nearby labels, prior notifications or grammatical choices created that inference. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 177: Research participation consent × Counterexample challenge

Reader-facing test. For a volunteer reading a study information sheet, begin with a consent form next to the ordinary sign-in process and do not disclose the intended interpretation. Invite the reader to invent an example in which the notice would not apply, or to say what evidence would make the current instruction no longer valid. The working question is whether the person can distinguish “voluntariness and past data handling are independent dimensions” from the tempting claim that declining means losing all access or withdrawal erases every existing record. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Inspect whether the reader recognizes limits, exceptions, unknown information and provisional states. The expected observation is that the person can explain both the freedom to decline and the limits of withdrawal. If instead the answer implies that declining means losing all access or withdrawal erases every existing record, log the exact words and where the reader looked. The potential harm is coerced participation or incorrect data expectations. Do not treat a fluent reformulation as proof of meaning; a single polite sentence may conceal two separate rights. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to state the independent services and the documented withdrawal consequences. An inability to imagine a counterexample may be a testing skill issue, so offer a neutral hypothetical rather than marking immediate failure. The key evidence is the reader’s understanding of the boundary after an ordinary scenario is supplied without revealing the answer. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 178: Research participation consent × Delayed teach-back

Reader-facing test. For a volunteer reading a study information sheet, begin with a consent form next to the ordinary sign-in process and do not disclose the intended interpretation. After an unrelated short filler activity, ask the reader to explain the next step to a fictional colleague who cannot view the translated message. The working question is whether the person can distinguish “voluntariness and past data handling are independent dimensions” from the tempting claim that declining means losing all access or withdrawal erases every existing record. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Note which conditions survive memory, which qualifiers disappear and whether confidence rises as detail is lost. The expected observation is that the person can explain both the freedom to decline and the limits of withdrawal. If instead the answer implies that declining means losing all access or withdrawal erases every existing record, log the exact words and where the reader looked. The potential harm is coerced participation or incorrect data expectations. Do not treat a fluent reformulation as proof of meaning; a single polite sentence may conceal two separate rights. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to state the independent services and the documented withdrawal consequences. Teach-back tests retained practical meaning, not memorization of exact wording. Compare it with first-exposure paraphrase. If the reader initially understood but later drops a critical exception, consider information placement, a reminder, a concise summary or a clearer call to action. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 179: Research participation consent × Alternative wording comparison

Reader-facing test. For a volunteer reading a study information sheet, begin with a consent form next to the ordinary sign-in process and do not disclose the intended interpretation. Show two approved candidate renderings in a balanced order across participants and ask what each would lead them to do; avoid asking which sounds more elegant. The working question is whether the person can distinguish “voluntariness and past data handling are independent dimensions” from the tempting claim that declining means losing all access or withdrawal erases every existing record. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Record whether the alternatives produce different real-world predictions despite similar fluency judgments. The expected observation is that the person can explain both the freedom to decline and the limits of withdrawal. If instead the answer implies that declining means losing all access or withdrawal erases every existing record, log the exact words and where the reader looked. The potential harm is coerced participation or incorrect data expectations. Do not treat a fluent reformulation as proof of meaning; a single polite sentence may conceal two separate rights. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to state the independent services and the documented withdrawal consequences. Do not optimize language by majority preference alone: a popular label can still induce a serious wrong decision. Preference is a supporting measure. Select wording that better preserves the rule, is natural for the intended audience and survives a subsequent task-based retest. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 180: Research participation consent × Release-readiness retest

Reader-facing test. For a volunteer reading a study information sheet, begin with a consent form next to the ordinary sign-in process and do not disclose the intended interpretation. Present the repaired artifact with the same surrounding interface but an independent fictional case; recruit someone not coached by the earlier test where feasible. The working question is whether the person can distinguish “voluntariness and past data handling are independent dimensions” from the tempting claim that declining means losing all access or withdrawal erases every existing record. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare unprompted decisions, explanations and remaining errors rather than merely tallying favorable reactions. The expected observation is that the person can explain both the freedom to decline and the limits of withdrawal. If instead the answer implies that declining means losing all access or withdrawal erases every existing record, log the exact words and where the reader looked. The potential harm is coerced participation or incorrect data expectations. Do not treat a fluent reformulation as proof of meaning; a single polite sentence may conceal two separate rights. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to state the independent services and the documented withdrawal consequences. A corrected phrase is not validated by the editor’s confidence. Preserve before-and-after versions, source approval, exact prompts, reader evidence and unresolved risks. Report that the tested failure diminished under these conditions, not that the translation is universally impossible to misunderstand. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Domain studio 21: Training certificate

A a learner reviewing completed-course status encounters a progress bar next to a certificate icon. The intended rule is: Completing the lessons records course participation; a separate assessed requirement must be satisfied before a certificate is issued. The risky alternative is finished lessons equals qualification obtained. This matters because unearned credential claims. Here the semantic hinge is that completion and competency recognition are different outcomes. These are invented teaching fixtures, so a real publisher must replace them with an approved source contract and its own audience research.

Diagnostic case 181: Training certificate × Blind paraphrase

Reader-facing test. For a learner reviewing completed-course status, begin with a progress bar next to a certificate icon and do not disclose the intended interpretation. Withhold the source original and ask the reader, in their own language, to explain the message as if describing it to someone who has not seen the screen. The working question is whether the person can distinguish “completion and competency recognition are different outcomes” from the tempting claim that finished lessons equals qualification obtained. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Listen for the causal story: what triggers the action, what stays true afterward and which actor controls the change. The expected observation is that the person can tell whether the certificate has actually been issued. If instead the answer implies that finished lessons equals qualification obtained, log the exact words and where the reader looked. The potential harm is unearned credential claims. Do not treat a fluent reformulation as proof of meaning; progress indicators are strong but fallible semantic cues. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to identify the assessment requirement beside the progress indicator. A clean paraphrase should preserve the decision rule rather than repeat isolated keywords. A polished but incorrect summary is more informative than a low-confidence guess because it shows the exact rule the reader has constructed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 182: Training certificate × Task simulation

Reader-facing test. For a learner reviewing completed-course status, begin with a progress bar next to a certificate icon and do not disclose the intended interpretation. Give a fictional scenario with a choice between acting now, waiting for confirmation and searching for another condition; allow the reader to use only the visible translated material. The working question is whether the person can distinguish “completion and competency recognition are different outcomes” from the tempting claim that finished lessons equals qualification obtained. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Capture the chosen action, the text used to justify it and whether a wrong branch seems attractive. The expected observation is that the person can tell whether the certificate has actually been issued. If instead the answer implies that finished lessons equals qualification obtained, log the exact words and where the reader looked. The potential harm is unearned credential claims. Do not treat a fluent reformulation as proof of meaning; progress indicators are strong but fallible semantic cues. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to identify the assessment requirement beside the progress indicator. Do not score the final click alone. Someone may guess correctly from placement while holding the wrong belief. A complete result includes the reason, expected consequence and what additional information the person believes is needed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 183: Training certificate × Think-aloud trace

Reader-facing test. For a learner reviewing completed-course status, begin with a progress bar next to a certificate icon and do not disclose the intended interpretation. Ask the reader to narrate thoughts while moving from the notice to the relevant action, but do not correct terminology or explain the organization’s policy. The working question is whether the person can distinguish “completion and competency recognition are different outcomes” from the tempting claim that finished lessons equals qualification obtained. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Mark the moment a term is interpreted, the moment confidence changes and the point at which layout or icons redirect attention. The expected observation is that the person can tell whether the certificate has actually been issued. If instead the answer implies that finished lessons equals qualification obtained, log the exact words and where the reader looked. The potential harm is unearned credential claims. Do not treat a fluent reformulation as proof of meaning; progress indicators are strong but fallible semantic cues. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to identify the assessment requirement beside the progress indicator. This method can identify an interface cue rather than a translation error. The very act of thinking aloud may change behavior, so report it as qualitative diagnostic evidence rather than a naturalistic time-to-completion measurement. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 184: Training certificate × Contrastive boundary pair

Reader-facing test. For a learner reviewing completed-course status, begin with a progress bar next to a certificate icon and do not disclose the intended interpretation. Present two nearly identical fictional cases that differ only in the decisive condition; ask whether the outcome changes and why. The working question is whether the person can distinguish “completion and competency recognition are different outcomes” from the tempting claim that finished lessons equals qualification obtained. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Look for the reader’s treatment of a status, exception or timing boundary that the source intended to preserve. The expected observation is that the person can tell whether the certificate has actually been issued. If instead the answer implies that finished lessons equals qualification obtained, log the exact words and where the reader looked. The potential harm is unearned credential claims. Do not treat a fluent reformulation as proof of meaning; progress indicators are strong but fallible semantic cues. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to identify the assessment requirement beside the progress indicator. A reader who gives the same answer to both cases may have lost the qualifier. However, if the source rule itself is ambiguous, pause the test and clarify it with the content owner. Do not force respondents to guess at a rule the author has not specified. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 185: Training certificate × Actor and responsibility map

Reader-facing test. For a learner reviewing completed-course status, begin with a progress bar next to a certificate icon and do not disclose the intended interpretation. Ask the reader to identify who submits, who decides, who can revise the decision and who is affected; let them point to the translated words supporting each answer. The working question is whether the person can distinguish “completion and competency recognition are different outcomes” from the tempting claim that finished lessons equals qualification obtained. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare the named actors with the interpretation contract and note any responsibility the reader invents. The expected observation is that the person can tell whether the certificate has actually been issued. If instead the answer implies that finished lessons equals qualification obtained, log the exact words and where the reader looked. The potential harm is unearned credential claims. Do not treat a fluent reformulation as proof of meaning; progress indicators are strong but fallible semantic cues. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to identify the assessment requirement beside the progress indicator. Passive voice is not automatically wrong, and some languages omit subjects naturally. The test is whether the actual actor is recoverable. If the reader supplies an unintended actor, identify whether nearby labels, prior notifications or grammatical choices created that inference. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 186: Training certificate × Counterexample challenge

Reader-facing test. For a learner reviewing completed-course status, begin with a progress bar next to a certificate icon and do not disclose the intended interpretation. Invite the reader to invent an example in which the notice would not apply, or to say what evidence would make the current instruction no longer valid. The working question is whether the person can distinguish “completion and competency recognition are different outcomes” from the tempting claim that finished lessons equals qualification obtained. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Inspect whether the reader recognizes limits, exceptions, unknown information and provisional states. The expected observation is that the person can tell whether the certificate has actually been issued. If instead the answer implies that finished lessons equals qualification obtained, log the exact words and where the reader looked. The potential harm is unearned credential claims. Do not treat a fluent reformulation as proof of meaning; progress indicators are strong but fallible semantic cues. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to identify the assessment requirement beside the progress indicator. An inability to imagine a counterexample may be a testing skill issue, so offer a neutral hypothetical rather than marking immediate failure. The key evidence is the reader’s understanding of the boundary after an ordinary scenario is supplied without revealing the answer. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 187: Training certificate × Delayed teach-back

Reader-facing test. For a learner reviewing completed-course status, begin with a progress bar next to a certificate icon and do not disclose the intended interpretation. After an unrelated short filler activity, ask the reader to explain the next step to a fictional colleague who cannot view the translated message. The working question is whether the person can distinguish “completion and competency recognition are different outcomes” from the tempting claim that finished lessons equals qualification obtained. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Note which conditions survive memory, which qualifiers disappear and whether confidence rises as detail is lost. The expected observation is that the person can tell whether the certificate has actually been issued. If instead the answer implies that finished lessons equals qualification obtained, log the exact words and where the reader looked. The potential harm is unearned credential claims. Do not treat a fluent reformulation as proof of meaning; progress indicators are strong but fallible semantic cues. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to identify the assessment requirement beside the progress indicator. Teach-back tests retained practical meaning, not memorization of exact wording. Compare it with first-exposure paraphrase. If the reader initially understood but later drops a critical exception, consider information placement, a reminder, a concise summary or a clearer call to action. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 188: Training certificate × Alternative wording comparison

Reader-facing test. For a learner reviewing completed-course status, begin with a progress bar next to a certificate icon and do not disclose the intended interpretation. Show two approved candidate renderings in a balanced order across participants and ask what each would lead them to do; avoid asking which sounds more elegant. The working question is whether the person can distinguish “completion and competency recognition are different outcomes” from the tempting claim that finished lessons equals qualification obtained. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Record whether the alternatives produce different real-world predictions despite similar fluency judgments. The expected observation is that the person can tell whether the certificate has actually been issued. If instead the answer implies that finished lessons equals qualification obtained, log the exact words and where the reader looked. The potential harm is unearned credential claims. Do not treat a fluent reformulation as proof of meaning; progress indicators are strong but fallible semantic cues. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to identify the assessment requirement beside the progress indicator. Do not optimize language by majority preference alone: a popular label can still induce a serious wrong decision. Preference is a supporting measure. Select wording that better preserves the rule, is natural for the intended audience and survives a subsequent task-based retest. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 189: Training certificate × Release-readiness retest

Reader-facing test. For a learner reviewing completed-course status, begin with a progress bar next to a certificate icon and do not disclose the intended interpretation. Present the repaired artifact with the same surrounding interface but an independent fictional case; recruit someone not coached by the earlier test where feasible. The working question is whether the person can distinguish “completion and competency recognition are different outcomes” from the tempting claim that finished lessons equals qualification obtained. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare unprompted decisions, explanations and remaining errors rather than merely tallying favorable reactions. The expected observation is that the person can tell whether the certificate has actually been issued. If instead the answer implies that finished lessons equals qualification obtained, log the exact words and where the reader looked. The potential harm is unearned credential claims. Do not treat a fluent reformulation as proof of meaning; progress indicators are strong but fallible semantic cues. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to identify the assessment requirement beside the progress indicator. A corrected phrase is not validated by the editor’s confidence. Preserve before-and-after versions, source approval, exact prompts, reader evidence and unresolved risks. Report that the tested failure diminished under these conditions, not that the translation is universally impossible to misunderstand. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Domain studio 22: Ticket transfer permission

A a ticket-holder sending access to another person encounters a share icon beside a transfer control. The intended rule is: Sharing a viewing link does not transfer ownership; a transfer is valid only after the designated transfer process confirms a new holder. The risky alternative is forwarding an email permanently transfers the ticket. This matters because denied entry or duplicate use. Here the semantic hinge is that possession of a link and registered ownership differ. These are invented teaching fixtures, so a real publisher must replace them with an approved source contract and its own audience research.

Diagnostic case 190: Ticket transfer permission × Blind paraphrase

Reader-facing test. For a ticket-holder sending access to another person, begin with a share icon beside a transfer control and do not disclose the intended interpretation. Withhold the source original and ask the reader, in their own language, to explain the message as if describing it to someone who has not seen the screen. The working question is whether the person can distinguish “possession of a link and registered ownership differ” from the tempting claim that forwarding an email permanently transfers the ticket. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Listen for the causal story: what triggers the action, what stays true afterward and which actor controls the change. The expected observation is that the person can identify who remains the registered holder after each action. If instead the answer implies that forwarding an email permanently transfers the ticket, log the exact words and where the reader looked. The potential harm is denied entry or duplicate use. Do not treat a fluent reformulation as proof of meaning; a familiar share verb can conceal identity state. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to separate preview-sharing from formal reassignment. A clean paraphrase should preserve the decision rule rather than repeat isolated keywords. A polished but incorrect summary is more informative than a low-confidence guess because it shows the exact rule the reader has constructed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 191: Ticket transfer permission × Task simulation

Reader-facing test. For a ticket-holder sending access to another person, begin with a share icon beside a transfer control and do not disclose the intended interpretation. Give a fictional scenario with a choice between acting now, waiting for confirmation and searching for another condition; allow the reader to use only the visible translated material. The working question is whether the person can distinguish “possession of a link and registered ownership differ” from the tempting claim that forwarding an email permanently transfers the ticket. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Capture the chosen action, the text used to justify it and whether a wrong branch seems attractive. The expected observation is that the person can identify who remains the registered holder after each action. If instead the answer implies that forwarding an email permanently transfers the ticket, log the exact words and where the reader looked. The potential harm is denied entry or duplicate use. Do not treat a fluent reformulation as proof of meaning; a familiar share verb can conceal identity state. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to separate preview-sharing from formal reassignment. Do not score the final click alone. Someone may guess correctly from placement while holding the wrong belief. A complete result includes the reason, expected consequence and what additional information the person believes is needed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 192: Ticket transfer permission × Think-aloud trace

Reader-facing test. For a ticket-holder sending access to another person, begin with a share icon beside a transfer control and do not disclose the intended interpretation. Ask the reader to narrate thoughts while moving from the notice to the relevant action, but do not correct terminology or explain the organization’s policy. The working question is whether the person can distinguish “possession of a link and registered ownership differ” from the tempting claim that forwarding an email permanently transfers the ticket. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Mark the moment a term is interpreted, the moment confidence changes and the point at which layout or icons redirect attention. The expected observation is that the person can identify who remains the registered holder after each action. If instead the answer implies that forwarding an email permanently transfers the ticket, log the exact words and where the reader looked. The potential harm is denied entry or duplicate use. Do not treat a fluent reformulation as proof of meaning; a familiar share verb can conceal identity state. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to separate preview-sharing from formal reassignment. This method can identify an interface cue rather than a translation error. The very act of thinking aloud may change behavior, so report it as qualitative diagnostic evidence rather than a naturalistic time-to-completion measurement. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 193: Ticket transfer permission × Contrastive boundary pair

Reader-facing test. For a ticket-holder sending access to another person, begin with a share icon beside a transfer control and do not disclose the intended interpretation. Present two nearly identical fictional cases that differ only in the decisive condition; ask whether the outcome changes and why. The working question is whether the person can distinguish “possession of a link and registered ownership differ” from the tempting claim that forwarding an email permanently transfers the ticket. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Look for the reader’s treatment of a status, exception or timing boundary that the source intended to preserve. The expected observation is that the person can identify who remains the registered holder after each action. If instead the answer implies that forwarding an email permanently transfers the ticket, log the exact words and where the reader looked. The potential harm is denied entry or duplicate use. Do not treat a fluent reformulation as proof of meaning; a familiar share verb can conceal identity state. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to separate preview-sharing from formal reassignment. A reader who gives the same answer to both cases may have lost the qualifier. However, if the source rule itself is ambiguous, pause the test and clarify it with the content owner. Do not force respondents to guess at a rule the author has not specified. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 194: Ticket transfer permission × Actor and responsibility map

Reader-facing test. For a ticket-holder sending access to another person, begin with a share icon beside a transfer control and do not disclose the intended interpretation. Ask the reader to identify who submits, who decides, who can revise the decision and who is affected; let them point to the translated words supporting each answer. The working question is whether the person can distinguish “possession of a link and registered ownership differ” from the tempting claim that forwarding an email permanently transfers the ticket. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare the named actors with the interpretation contract and note any responsibility the reader invents. The expected observation is that the person can identify who remains the registered holder after each action. If instead the answer implies that forwarding an email permanently transfers the ticket, log the exact words and where the reader looked. The potential harm is denied entry or duplicate use. Do not treat a fluent reformulation as proof of meaning; a familiar share verb can conceal identity state. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to separate preview-sharing from formal reassignment. Passive voice is not automatically wrong, and some languages omit subjects naturally. The test is whether the actual actor is recoverable. If the reader supplies an unintended actor, identify whether nearby labels, prior notifications or grammatical choices created that inference. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 195: Ticket transfer permission × Counterexample challenge

Reader-facing test. For a ticket-holder sending access to another person, begin with a share icon beside a transfer control and do not disclose the intended interpretation. Invite the reader to invent an example in which the notice would not apply, or to say what evidence would make the current instruction no longer valid. The working question is whether the person can distinguish “possession of a link and registered ownership differ” from the tempting claim that forwarding an email permanently transfers the ticket. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Inspect whether the reader recognizes limits, exceptions, unknown information and provisional states. The expected observation is that the person can identify who remains the registered holder after each action. If instead the answer implies that forwarding an email permanently transfers the ticket, log the exact words and where the reader looked. The potential harm is denied entry or duplicate use. Do not treat a fluent reformulation as proof of meaning; a familiar share verb can conceal identity state. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to separate preview-sharing from formal reassignment. An inability to imagine a counterexample may be a testing skill issue, so offer a neutral hypothetical rather than marking immediate failure. The key evidence is the reader’s understanding of the boundary after an ordinary scenario is supplied without revealing the answer. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 196: Ticket transfer permission × Delayed teach-back

Reader-facing test. For a ticket-holder sending access to another person, begin with a share icon beside a transfer control and do not disclose the intended interpretation. After an unrelated short filler activity, ask the reader to explain the next step to a fictional colleague who cannot view the translated message. The working question is whether the person can distinguish “possession of a link and registered ownership differ” from the tempting claim that forwarding an email permanently transfers the ticket. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Note which conditions survive memory, which qualifiers disappear and whether confidence rises as detail is lost. The expected observation is that the person can identify who remains the registered holder after each action. If instead the answer implies that forwarding an email permanently transfers the ticket, log the exact words and where the reader looked. The potential harm is denied entry or duplicate use. Do not treat a fluent reformulation as proof of meaning; a familiar share verb can conceal identity state. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to separate preview-sharing from formal reassignment. Teach-back tests retained practical meaning, not memorization of exact wording. Compare it with first-exposure paraphrase. If the reader initially understood but later drops a critical exception, consider information placement, a reminder, a concise summary or a clearer call to action. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 197: Ticket transfer permission × Alternative wording comparison

Reader-facing test. For a ticket-holder sending access to another person, begin with a share icon beside a transfer control and do not disclose the intended interpretation. Show two approved candidate renderings in a balanced order across participants and ask what each would lead them to do; avoid asking which sounds more elegant. The working question is whether the person can distinguish “possession of a link and registered ownership differ” from the tempting claim that forwarding an email permanently transfers the ticket. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Record whether the alternatives produce different real-world predictions despite similar fluency judgments. The expected observation is that the person can identify who remains the registered holder after each action. If instead the answer implies that forwarding an email permanently transfers the ticket, log the exact words and where the reader looked. The potential harm is denied entry or duplicate use. Do not treat a fluent reformulation as proof of meaning; a familiar share verb can conceal identity state. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to separate preview-sharing from formal reassignment. Do not optimize language by majority preference alone: a popular label can still induce a serious wrong decision. Preference is a supporting measure. Select wording that better preserves the rule, is natural for the intended audience and survives a subsequent task-based retest. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 198: Ticket transfer permission × Release-readiness retest

Reader-facing test. For a ticket-holder sending access to another person, begin with a share icon beside a transfer control and do not disclose the intended interpretation. Present the repaired artifact with the same surrounding interface but an independent fictional case; recruit someone not coached by the earlier test where feasible. The working question is whether the person can distinguish “possession of a link and registered ownership differ” from the tempting claim that forwarding an email permanently transfers the ticket. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare unprompted decisions, explanations and remaining errors rather than merely tallying favorable reactions. The expected observation is that the person can identify who remains the registered holder after each action. If instead the answer implies that forwarding an email permanently transfers the ticket, log the exact words and where the reader looked. The potential harm is denied entry or duplicate use. Do not treat a fluent reformulation as proof of meaning; a familiar share verb can conceal identity state. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to separate preview-sharing from formal reassignment. A corrected phrase is not validated by the editor’s confidence. Preserve before-and-after versions, source approval, exact prompts, reader evidence and unresolved risks. Report that the tested failure diminished under these conditions, not that the translation is universally impossible to misunderstand. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Domain studio 23: Software security update banner

A a user responding to an application update encounters a download notification next to an install-status indicator. The intended rule is: A downloaded update is not necessarily installed; a restart or completion signal may still be required before the new version is active. The risky alternative is downloaded equals protected by the new version. This matters because false belief that an update is active. Here the semantic hinge is that file acquisition and successful deployment are separate states. These are invented teaching fixtures, so a real publisher must replace them with an approved source contract and its own audience research.

Diagnostic case 199: Software security update banner × Blind paraphrase

Reader-facing test. For a user responding to an application update, begin with a download notification next to an install-status indicator and do not disclose the intended interpretation. Withhold the source original and ask the reader, in their own language, to explain the message as if describing it to someone who has not seen the screen. The working question is whether the person can distinguish “file acquisition and successful deployment are separate states” from the tempting claim that downloaded equals protected by the new version. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Listen for the causal story: what triggers the action, what stays true afterward and which actor controls the change. The expected observation is that the person can identify which stage is complete on the shown screen. If instead the answer implies that downloaded equals protected by the new version, log the exact words and where the reader looked. The potential harm is false belief that an update is active. Do not treat a fluent reformulation as proof of meaning; technical progress terms often collapse into one everyday verb. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to describe downloaded, installed and active states distinctly. A clean paraphrase should preserve the decision rule rather than repeat isolated keywords. A polished but incorrect summary is more informative than a low-confidence guess because it shows the exact rule the reader has constructed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 200: Software security update banner × Task simulation

Reader-facing test. For a user responding to an application update, begin with a download notification next to an install-status indicator and do not disclose the intended interpretation. Give a fictional scenario with a choice between acting now, waiting for confirmation and searching for another condition; allow the reader to use only the visible translated material. The working question is whether the person can distinguish “file acquisition and successful deployment are separate states” from the tempting claim that downloaded equals protected by the new version. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Capture the chosen action, the text used to justify it and whether a wrong branch seems attractive. The expected observation is that the person can identify which stage is complete on the shown screen. If instead the answer implies that downloaded equals protected by the new version, log the exact words and where the reader looked. The potential harm is false belief that an update is active. Do not treat a fluent reformulation as proof of meaning; technical progress terms often collapse into one everyday verb. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to describe downloaded, installed and active states distinctly. Do not score the final click alone. Someone may guess correctly from placement while holding the wrong belief. A complete result includes the reason, expected consequence and what additional information the person believes is needed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 201: Software security update banner × Think-aloud trace

Reader-facing test. For a user responding to an application update, begin with a download notification next to an install-status indicator and do not disclose the intended interpretation. Ask the reader to narrate thoughts while moving from the notice to the relevant action, but do not correct terminology or explain the organization’s policy. The working question is whether the person can distinguish “file acquisition and successful deployment are separate states” from the tempting claim that downloaded equals protected by the new version. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Mark the moment a term is interpreted, the moment confidence changes and the point at which layout or icons redirect attention. The expected observation is that the person can identify which stage is complete on the shown screen. If instead the answer implies that downloaded equals protected by the new version, log the exact words and where the reader looked. The potential harm is false belief that an update is active. Do not treat a fluent reformulation as proof of meaning; technical progress terms often collapse into one everyday verb. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to describe downloaded, installed and active states distinctly. This method can identify an interface cue rather than a translation error. The very act of thinking aloud may change behavior, so report it as qualitative diagnostic evidence rather than a naturalistic time-to-completion measurement. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 202: Software security update banner × Contrastive boundary pair

Reader-facing test. For a user responding to an application update, begin with a download notification next to an install-status indicator and do not disclose the intended interpretation. Present two nearly identical fictional cases that differ only in the decisive condition; ask whether the outcome changes and why. The working question is whether the person can distinguish “file acquisition and successful deployment are separate states” from the tempting claim that downloaded equals protected by the new version. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Look for the reader’s treatment of a status, exception or timing boundary that the source intended to preserve. The expected observation is that the person can identify which stage is complete on the shown screen. If instead the answer implies that downloaded equals protected by the new version, log the exact words and where the reader looked. The potential harm is false belief that an update is active. Do not treat a fluent reformulation as proof of meaning; technical progress terms often collapse into one everyday verb. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to describe downloaded, installed and active states distinctly. A reader who gives the same answer to both cases may have lost the qualifier. However, if the source rule itself is ambiguous, pause the test and clarify it with the content owner. Do not force respondents to guess at a rule the author has not specified. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 203: Software security update banner × Actor and responsibility map

Reader-facing test. For a user responding to an application update, begin with a download notification next to an install-status indicator and do not disclose the intended interpretation. Ask the reader to identify who submits, who decides, who can revise the decision and who is affected; let them point to the translated words supporting each answer. The working question is whether the person can distinguish “file acquisition and successful deployment are separate states” from the tempting claim that downloaded equals protected by the new version. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare the named actors with the interpretation contract and note any responsibility the reader invents. The expected observation is that the person can identify which stage is complete on the shown screen. If instead the answer implies that downloaded equals protected by the new version, log the exact words and where the reader looked. The potential harm is false belief that an update is active. Do not treat a fluent reformulation as proof of meaning; technical progress terms often collapse into one everyday verb. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to describe downloaded, installed and active states distinctly. Passive voice is not automatically wrong, and some languages omit subjects naturally. The test is whether the actual actor is recoverable. If the reader supplies an unintended actor, identify whether nearby labels, prior notifications or grammatical choices created that inference. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 204: Software security update banner × Counterexample challenge

Reader-facing test. For a user responding to an application update, begin with a download notification next to an install-status indicator and do not disclose the intended interpretation. Invite the reader to invent an example in which the notice would not apply, or to say what evidence would make the current instruction no longer valid. The working question is whether the person can distinguish “file acquisition and successful deployment are separate states” from the tempting claim that downloaded equals protected by the new version. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Inspect whether the reader recognizes limits, exceptions, unknown information and provisional states. The expected observation is that the person can identify which stage is complete on the shown screen. If instead the answer implies that downloaded equals protected by the new version, log the exact words and where the reader looked. The potential harm is false belief that an update is active. Do not treat a fluent reformulation as proof of meaning; technical progress terms often collapse into one everyday verb. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to describe downloaded, installed and active states distinctly. An inability to imagine a counterexample may be a testing skill issue, so offer a neutral hypothetical rather than marking immediate failure. The key evidence is the reader’s understanding of the boundary after an ordinary scenario is supplied without revealing the answer. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 205: Software security update banner × Delayed teach-back

Reader-facing test. For a user responding to an application update, begin with a download notification next to an install-status indicator and do not disclose the intended interpretation. After an unrelated short filler activity, ask the reader to explain the next step to a fictional colleague who cannot view the translated message. The working question is whether the person can distinguish “file acquisition and successful deployment are separate states” from the tempting claim that downloaded equals protected by the new version. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Note which conditions survive memory, which qualifiers disappear and whether confidence rises as detail is lost. The expected observation is that the person can identify which stage is complete on the shown screen. If instead the answer implies that downloaded equals protected by the new version, log the exact words and where the reader looked. The potential harm is false belief that an update is active. Do not treat a fluent reformulation as proof of meaning; technical progress terms often collapse into one everyday verb. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to describe downloaded, installed and active states distinctly. Teach-back tests retained practical meaning, not memorization of exact wording. Compare it with first-exposure paraphrase. If the reader initially understood but later drops a critical exception, consider information placement, a reminder, a concise summary or a clearer call to action. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 206: Software security update banner × Alternative wording comparison

Reader-facing test. For a user responding to an application update, begin with a download notification next to an install-status indicator and do not disclose the intended interpretation. Show two approved candidate renderings in a balanced order across participants and ask what each would lead them to do; avoid asking which sounds more elegant. The working question is whether the person can distinguish “file acquisition and successful deployment are separate states” from the tempting claim that downloaded equals protected by the new version. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Record whether the alternatives produce different real-world predictions despite similar fluency judgments. The expected observation is that the person can identify which stage is complete on the shown screen. If instead the answer implies that downloaded equals protected by the new version, log the exact words and where the reader looked. The potential harm is false belief that an update is active. Do not treat a fluent reformulation as proof of meaning; technical progress terms often collapse into one everyday verb. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to describe downloaded, installed and active states distinctly. Do not optimize language by majority preference alone: a popular label can still induce a serious wrong decision. Preference is a supporting measure. Select wording that better preserves the rule, is natural for the intended audience and survives a subsequent task-based retest. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 207: Software security update banner × Release-readiness retest

Reader-facing test. For a user responding to an application update, begin with a download notification next to an install-status indicator and do not disclose the intended interpretation. Present the repaired artifact with the same surrounding interface but an independent fictional case; recruit someone not coached by the earlier test where feasible. The working question is whether the person can distinguish “file acquisition and successful deployment are separate states” from the tempting claim that downloaded equals protected by the new version. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare unprompted decisions, explanations and remaining errors rather than merely tallying favorable reactions. The expected observation is that the person can identify which stage is complete on the shown screen. If instead the answer implies that downloaded equals protected by the new version, log the exact words and where the reader looked. The potential harm is false belief that an update is active. Do not treat a fluent reformulation as proof of meaning; technical progress terms often collapse into one everyday verb. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to describe downloaded, installed and active states distinctly. A corrected phrase is not validated by the editor’s confidence. Preserve before-and-after versions, source approval, exact prompts, reader evidence and unresolved risks. Report that the tested failure diminished under these conditions, not that the translation is universally impossible to misunderstand. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Domain studio 24: Community assistance request

A a resident completing a request for non-emergency help encounters a request receipt and reference number. The intended rule is: Submitting an assistance request confirms receipt, while assessment, eligibility and service allocation happen later. The risky alternative is request received guarantees help at a specified time. This matters because false reassurance and missed alternative options. Here the semantic hinge is that intake confirmation is not an allocation promise. These are invented teaching fixtures, so a real publisher must replace them with an approved source contract and its own audience research.

Diagnostic case 208: Community assistance request × Blind paraphrase

Reader-facing test. For a resident completing a request for non-emergency help, begin with a request receipt and reference number and do not disclose the intended interpretation. Withhold the source original and ask the reader, in their own language, to explain the message as if describing it to someone who has not seen the screen. The working question is whether the person can distinguish “intake confirmation is not an allocation promise” from the tempting claim that request received guarantees help at a specified time. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Listen for the causal story: what triggers the action, what stays true afterward and which actor controls the change. The expected observation is that the person can explain the next process without inventing an appointment. If instead the answer implies that request received guarantees help at a specified time, log the exact words and where the reader looked. The potential harm is false reassurance and missed alternative options. Do not treat a fluent reformulation as proof of meaning; empathy in wording must not fabricate a commitment. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label the response status and explain what the receipt proves. A clean paraphrase should preserve the decision rule rather than repeat isolated keywords. A polished but incorrect summary is more informative than a low-confidence guess because it shows the exact rule the reader has constructed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 209: Community assistance request × Task simulation

Reader-facing test. For a resident completing a request for non-emergency help, begin with a request receipt and reference number and do not disclose the intended interpretation. Give a fictional scenario with a choice between acting now, waiting for confirmation and searching for another condition; allow the reader to use only the visible translated material. The working question is whether the person can distinguish “intake confirmation is not an allocation promise” from the tempting claim that request received guarantees help at a specified time. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Capture the chosen action, the text used to justify it and whether a wrong branch seems attractive. The expected observation is that the person can explain the next process without inventing an appointment. If instead the answer implies that request received guarantees help at a specified time, log the exact words and where the reader looked. The potential harm is false reassurance and missed alternative options. Do not treat a fluent reformulation as proof of meaning; empathy in wording must not fabricate a commitment. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label the response status and explain what the receipt proves. Do not score the final click alone. Someone may guess correctly from placement while holding the wrong belief. A complete result includes the reason, expected consequence and what additional information the person believes is needed. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 210: Community assistance request × Think-aloud trace

Reader-facing test. For a resident completing a request for non-emergency help, begin with a request receipt and reference number and do not disclose the intended interpretation. Ask the reader to narrate thoughts while moving from the notice to the relevant action, but do not correct terminology or explain the organization’s policy. The working question is whether the person can distinguish “intake confirmation is not an allocation promise” from the tempting claim that request received guarantees help at a specified time. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Mark the moment a term is interpreted, the moment confidence changes and the point at which layout or icons redirect attention. The expected observation is that the person can explain the next process without inventing an appointment. If instead the answer implies that request received guarantees help at a specified time, log the exact words and where the reader looked. The potential harm is false reassurance and missed alternative options. Do not treat a fluent reformulation as proof of meaning; empathy in wording must not fabricate a commitment. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label the response status and explain what the receipt proves. This method can identify an interface cue rather than a translation error. The very act of thinking aloud may change behavior, so report it as qualitative diagnostic evidence rather than a naturalistic time-to-completion measurement. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 211: Community assistance request × Contrastive boundary pair

Reader-facing test. For a resident completing a request for non-emergency help, begin with a request receipt and reference number and do not disclose the intended interpretation. Present two nearly identical fictional cases that differ only in the decisive condition; ask whether the outcome changes and why. The working question is whether the person can distinguish “intake confirmation is not an allocation promise” from the tempting claim that request received guarantees help at a specified time. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Look for the reader’s treatment of a status, exception or timing boundary that the source intended to preserve. The expected observation is that the person can explain the next process without inventing an appointment. If instead the answer implies that request received guarantees help at a specified time, log the exact words and where the reader looked. The potential harm is false reassurance and missed alternative options. Do not treat a fluent reformulation as proof of meaning; empathy in wording must not fabricate a commitment. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label the response status and explain what the receipt proves. A reader who gives the same answer to both cases may have lost the qualifier. However, if the source rule itself is ambiguous, pause the test and clarify it with the content owner. Do not force respondents to guess at a rule the author has not specified. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 212: Community assistance request × Actor and responsibility map

Reader-facing test. For a resident completing a request for non-emergency help, begin with a request receipt and reference number and do not disclose the intended interpretation. Ask the reader to identify who submits, who decides, who can revise the decision and who is affected; let them point to the translated words supporting each answer. The working question is whether the person can distinguish “intake confirmation is not an allocation promise” from the tempting claim that request received guarantees help at a specified time. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare the named actors with the interpretation contract and note any responsibility the reader invents. The expected observation is that the person can explain the next process without inventing an appointment. If instead the answer implies that request received guarantees help at a specified time, log the exact words and where the reader looked. The potential harm is false reassurance and missed alternative options. Do not treat a fluent reformulation as proof of meaning; empathy in wording must not fabricate a commitment. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label the response status and explain what the receipt proves. Passive voice is not automatically wrong, and some languages omit subjects naturally. The test is whether the actual actor is recoverable. If the reader supplies an unintended actor, identify whether nearby labels, prior notifications or grammatical choices created that inference. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 213: Community assistance request × Counterexample challenge

Reader-facing test. For a resident completing a request for non-emergency help, begin with a request receipt and reference number and do not disclose the intended interpretation. Invite the reader to invent an example in which the notice would not apply, or to say what evidence would make the current instruction no longer valid. The working question is whether the person can distinguish “intake confirmation is not an allocation promise” from the tempting claim that request received guarantees help at a specified time. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Inspect whether the reader recognizes limits, exceptions, unknown information and provisional states. The expected observation is that the person can explain the next process without inventing an appointment. If instead the answer implies that request received guarantees help at a specified time, log the exact words and where the reader looked. The potential harm is false reassurance and missed alternative options. Do not treat a fluent reformulation as proof of meaning; empathy in wording must not fabricate a commitment. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label the response status and explain what the receipt proves. An inability to imagine a counterexample may be a testing skill issue, so offer a neutral hypothetical rather than marking immediate failure. The key evidence is the reader’s understanding of the boundary after an ordinary scenario is supplied without revealing the answer. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 214: Community assistance request × Delayed teach-back

Reader-facing test. For a resident completing a request for non-emergency help, begin with a request receipt and reference number and do not disclose the intended interpretation. After an unrelated short filler activity, ask the reader to explain the next step to a fictional colleague who cannot view the translated message. The working question is whether the person can distinguish “intake confirmation is not an allocation promise” from the tempting claim that request received guarantees help at a specified time. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Note which conditions survive memory, which qualifiers disappear and whether confidence rises as detail is lost. The expected observation is that the person can explain the next process without inventing an appointment. If instead the answer implies that request received guarantees help at a specified time, log the exact words and where the reader looked. The potential harm is false reassurance and missed alternative options. Do not treat a fluent reformulation as proof of meaning; empathy in wording must not fabricate a commitment. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label the response status and explain what the receipt proves. Teach-back tests retained practical meaning, not memorization of exact wording. Compare it with first-exposure paraphrase. If the reader initially understood but later drops a critical exception, consider information placement, a reminder, a concise summary or a clearer call to action. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 215: Community assistance request × Alternative wording comparison

Reader-facing test. For a resident completing a request for non-emergency help, begin with a request receipt and reference number and do not disclose the intended interpretation. Show two approved candidate renderings in a balanced order across participants and ask what each would lead them to do; avoid asking which sounds more elegant. The working question is whether the person can distinguish “intake confirmation is not an allocation promise” from the tempting claim that request received guarantees help at a specified time. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Record whether the alternatives produce different real-world predictions despite similar fluency judgments. The expected observation is that the person can explain the next process without inventing an appointment. If instead the answer implies that request received guarantees help at a specified time, log the exact words and where the reader looked. The potential harm is false reassurance and missed alternative options. Do not treat a fluent reformulation as proof of meaning; empathy in wording must not fabricate a commitment. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label the response status and explain what the receipt proves. Do not optimize language by majority preference alone: a popular label can still induce a serious wrong decision. Preference is a supporting measure. Select wording that better preserves the rule, is natural for the intended audience and survives a subsequent task-based retest. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Diagnostic case 216: Community assistance request × Release-readiness retest

Reader-facing test. For a resident completing a request for non-emergency help, begin with a request receipt and reference number and do not disclose the intended interpretation. Present the repaired artifact with the same surrounding interface but an independent fictional case; recruit someone not coached by the earlier test where feasible. The working question is whether the person can distinguish “intake confirmation is not an allocation promise” from the tempting claim that request received guarantees help at a specified time. Make the decision consequential within a fictional task, but never ask for real sensitive details or a real transaction. If a facilitator names the correct rule first, the session no longer measures independent initial comprehension.

Observed evidence and diagnosis. Compare unprompted decisions, explanations and remaining errors rather than merely tallying favorable reactions. The expected observation is that the person can explain the next process without inventing an appointment. If instead the answer implies that request received guarantees help at a specified time, log the exact words and where the reader looked. The potential harm is false reassurance and missed alternative options. Do not treat a fluent reformulation as proof of meaning; empathy in wording must not fabricate a commitment. The analyst should determine whether the failure originates in the source, translation, label, context or unstated organizational rule before assigning responsibility.

Repair, transfer and retest. A plausible correction is to label the response status and explain what the receipt proves. A corrected phrase is not validated by the editor’s confidence. Preserve before-and-after versions, source approval, exact prompts, reader evidence and unresolved risks. Report that the tested failure diminished under these conditions, not that the translation is universally impossible to misunderstand. Use a new, comparable scenario to test that the repair works without the original interviewer’s cues. Keep the source version and the negative observation in the record, since an apparent success can conceal a new interpretation error elsewhere in the flow. In reporting, distinguish an actual participant response from the hypothetical failure illustrated in this teaching case.

Part IV. Twelve advanced diagnostic workshops

A. Distinguishing misunderstanding from disagreement

Suppose a reader correctly understands that a fee is nonrefundable but objects to the rule. A researcher who asks whether the person likes the wording may classify the objection as confusion. Instead, ask the reader what the rule requires, then separately ask whether they consider it fair. Record comprehension and preference in different fields. If the rule is understood yet disliked, changing the translation to make the fee sound softer is not an honest fix. If the wording induces a belief that the fee can be refunded under an unstated condition, that is a semantic defect. A truthful translation can be both disliked and successful. The study must avoid pushing participants toward the answer the organization wishes were popular.

Practical diagnostic record: note the source baseline, the exact target screen or notice, the participant’s unprompted interpretation, the action taken, an alternative scenario that changes one variable, the probable mechanism, the approved correction and the scope of the retest. A reviewer should be able to recreate the reasoning without relying on a verbal assurance that the session ‘went well.’

B. Distinguishing low vocabulary from missing context

A respondent may not recognize a specialist term. Offer no definition during the first-exposure measure; ask what they believe it refers to and which nearby information they rely on. In a separate diagnostic phase, give a neutral definition approved by the domain owner and ask the task again. Improvement after explanation supports a knowledge barrier, while continued failure suggests a structural problem or misleading relation among concepts. This does not mean replacing every technical term with an imprecise synonym. An effective repair can keep the required term and add a short explanation at the moment of use. Preserve the terminology owner’s definitions and avoid accidentally moving a vocabulary lesson into an unrelated translation quality decision.

Practical diagnostic record: note the source baseline, the exact target screen or notice, the participant’s unprompted interpretation, the action taken, an alternative scenario that changes one variable, the probable mechanism, the approved correction and the scope of the retest. A reviewer should be able to recreate the reasoning without relying on a verbal assurance that the session ‘went well.’

C. Distinguishing interface failure from linguistic failure

The translated rule may be explicit in a paragraph but invisible below a collapsed panel. A participant who never opens the panel cannot be counted as having understood or misunderstood its contents; the immediate failure is discovery. Compare two realistic displays using identical approved wording. If participants find the condition in one and not the other, the design is part of the cause. If they find it in both but interpret it incorrectly, language or domain knowledge may be implicated. Resist rewriting an accurate clause merely to conceal a navigation problem. Record both exposure and interpretation in a task trace.

Practical diagnostic record: note the source baseline, the exact target screen or notice, the participant’s unprompted interpretation, the action taken, an alternative scenario that changes one variable, the probable mechanism, the approved correction and the scope of the retest. A reviewer should be able to recreate the reasoning without relying on a verbal assurance that the session ‘went well.’

D. Constructing a threshold test

Consider a notice that offers a choice until a deadline. First freeze what the deadline means: local calendar day, exact timestamp, processing cutoff or receipt confirmation. Create three fictional cases before, at and after the boundary. Ask for the allowed action in each, not just the meaning of the word ‘until’. If the source owner cannot explain the at-boundary case, the issue begins in the source. If the target text appears to move the condition, document the translation effect. Never substitute a universal business-day convention when the publisher has defined something else.

Practical diagnostic record: note the source baseline, the exact target screen or notice, the participant’s unprompted interpretation, the action taken, an alternative scenario that changes one variable, the probable mechanism, the approved correction and the scope of the retest. A reviewer should be able to recreate the reasoning without relying on a verbal assurance that the session ‘went well.’

E. Testing uncertainty and probability

A service status may say a result is expected rather than guaranteed. Readers commonly convert reassuring language into certainty. Give a scenario in which the expected event does not occur and ask whether that would contradict the message. If the reader believes an estimate is a promise, determine whether the target rendering added a categorical verb or whether the surrounding visual presentation falsely implies finality. Avoid using strong statistical claims without numbers or source support. Test realistic choices under uncertainty, not a person’s appetite for risk.

Practical diagnostic record: note the source baseline, the exact target screen or notice, the participant’s unprompted interpretation, the action taken, an alternative scenario that changes one variable, the probable mechanism, the approved correction and the scope of the retest. A reviewer should be able to recreate the reasoning without relying on a verbal assurance that the session ‘went well.’

F. Comparing novice and expert readers

A specialist recognizes shorthand that a novice treats as ordinary language. Recruit both, without suggesting that the expert is always correct about how a public message should be read. Test the same scenario first and ask each to justify the decision. If only experts can infer the essential rule, the public-facing content may require an explanation, even when the specialist terminology is formally correct. Conversely, a public summary must not replace precise professional language where that precision is required. Design a layered communication approach and test each intended readership.

Practical diagnostic record: note the source baseline, the exact target screen or notice, the participant’s unprompted interpretation, the action taken, an alternative scenario that changes one variable, the probable mechanism, the approved correction and the scope of the retest. A reviewer should be able to recreate the reasoning without relying on a verbal assurance that the session ‘went well.’

G. Testing multilingual mixed-script displays

A string may contain an account reference, date, numerical interval or foreign product name embedded in a different writing direction. A participant may read the words yet associate the number with the wrong label because the visual order changes. Test on real device layouts with the approved language metadata; note spoken and visual reading order separately. When debugging, keep the literal identifier unchanged while isolating the layout and direction mechanism. This is a specialized transfer from comprehension study into established language-direction and accessibility work, not a license to change the underlying account data.

Practical diagnostic record: note the source baseline, the exact target screen or notice, the participant’s unprompted interpretation, the action taken, an alternative scenario that changes one variable, the probable mechanism, the approved correction and the scope of the retest. A reviewer should be able to recreate the reasoning without relying on a verbal assurance that the session ‘went well.’

H. Interviewer neutrality under pressure

A nervous respondent may ask the moderator, ‘Is this correct?’ Before the session, prepare a response such as ‘Please choose what you would do if you were using this by yourself.’ Do not encourage the respondent toward the desired behavior with nods, timing hints or recaps of the correct sentence. The moderator can ethically stop an exercise that causes distress while recording that it was incomplete. Staff who know the source language should be especially careful not to paraphrase in a way that supplies the intended answer. Audit prompts alongside results.

Practical diagnostic record: note the source baseline, the exact target screen or notice, the participant’s unprompted interpretation, the action taken, an alternative scenario that changes one variable, the probable mechanism, the approved correction and the scope of the retest. A reviewer should be able to recreate the reasoning without relying on a verbal assurance that the session ‘went well.’

I. Working with a multilingual team

When analysts do not share the participant’s language, have a qualified moderator capture the original response and a separate reviewer produce an evidence-preserving translation of key comments. Keep both versions linked to the same timestamp. Do not rely on an enthusiastic English summary that turns hesitation into confident comprehension. Distinguish participant paraphrase from analyst interpretation; disagreements should be adjudicated against the source contract and full transcript. Protect personal data and apply consent procedures appropriate to the research activity.

Practical diagnostic record: note the source baseline, the exact target screen or notice, the participant’s unprompted interpretation, the action taken, an alternative scenario that changes one variable, the probable mechanism, the approved correction and the scope of the retest. A reviewer should be able to recreate the reasoning without relying on a verbal assurance that the session ‘went well.’

J. Low-frequency but severe misunderstanding

In a small study, no observed failures do not establish that a rare and consequential misreading is absent. Use designed negative cases, expert hazard analysis and targeted boundary probes in addition to ordinary observation. A risk with severe consequences may justify stronger review and controlled deployment even when participant comments are favorable. Record what was not tested: device types, literacy profiles, unusual states, overlapping alerts or uncommon exceptions. The right conclusion is conditional: the tested version supported the intended interpretation in the observed scenarios, while additional assurance remains necessary.

Practical diagnostic record: note the source baseline, the exact target screen or notice, the participant’s unprompted interpretation, the action taken, an alternative scenario that changes one variable, the probable mechanism, the approved correction and the scope of the retest. A reviewer should be able to recreate the reasoning without relying on a verbal assurance that the session ‘went well.’

K. Handling disagreement about the ‘correct’ interpretation

Sometimes two source owners disagree whether a request status counts as approval. That is not a translation problem to solve through a reader poll. Freeze the release decision, obtain an authoritative clarification, update the source artifact and then re-examine each language version. The test may have exposed institutional ambiguity that already existed. Preserve evidence of the reader’s original inference without treating their answer as a vote on policy. A clean chain of accountability prevents a target-language editor from silently becoming the policy author.

Practical diagnostic record: note the source baseline, the exact target screen or notice, the participant’s unprompted interpretation, the action taken, an alternative scenario that changes one variable, the probable mechanism, the approved correction and the scope of the retest. A reviewer should be able to recreate the reasoning without relying on a verbal assurance that the session ‘went well.’

L. Writing an honest result report

Begin with the question tested, the intended audience and the exact artifact version. Include task scenarios, language variations, recruitment exclusions and moderator scripts. Present observable errors with quotations or screenshots as authorized; separate them from analyst hypotheses. Show what changed and what independent retest evidence exists. Do not label an entire language successful on the basis of a few tasks, and do not present invented illustrative cases as real findings. Conclude with a release recommendation, unresolved limitations and named owners for further investigation. This makes comprehension testing a repeatable engineering activity rather than a certificate printed from a satisfaction survey.

Practical diagnostic record: note the source baseline, the exact target screen or notice, the participant’s unprompted interpretation, the action taken, an alternative scenario that changes one variable, the probable mechanism, the approved correction and the scope of the retest. A reviewer should be able to recreate the reasoning without relying on a verbal assurance that the session ‘went well.’

Part V. Failure modes and the questions that expose them

Failure 1: Leading questions

Detect: Did the moderator leak a key word or the correct answer before the first interpretation? Resolve: Compare responses before and after intervention and score only the appropriate stage. The analyst should compare the actual evidence with the frozen intention, record the smallest plausible root cause and select a revision that does not introduce a new obligation. If no participant evidence exists, describe the issue as a predicted risk or editorial hypothesis rather than an observed reader failure.

Failure 2: Perfect grammar, wrong consequence

Detect: Does a natural sentence imply a different action or policy effect? Resolve: Trace the reader’s predicted outcome, not a grammar rating. The analyst should compare the actual evidence with the frozen intention, record the smallest plausible root cause and select a revision that does not introduce a new obligation. If no participant evidence exists, describe the issue as a predicted risk or editorial hypothesis rather than an observed reader failure.

Failure 3: Source ambiguity

Detect: Can the source owner state an unambiguous intended rule for the boundary case? Resolve: Clarify the source and propagate approved changes to the target versions. The analyst should compare the actual evidence with the frozen intention, record the smallest plausible root cause and select a revision that does not introduce a new obligation. If no participant evidence exists, describe the issue as a predicted risk or editorial hypothesis rather than an observed reader failure.

Failure 4: Word familiarity trap

Detect: Did a familiar term activate an unrelated institutional procedure? Resolve: Ask for a prediction of what happens next and identify imported assumptions. The analyst should compare the actual evidence with the frozen intention, record the smallest plausible root cause and select a revision that does not introduce a new obligation. If no participant evidence exists, describe the issue as a predicted risk or editorial hypothesis rather than an observed reader failure.

Failure 5: Unseen text

Detect: Was the qualifying sentence visible at the point of choice? Resolve: Measure exposure and discovery before interpreting a failure as linguistic. The analyst should compare the actual evidence with the frozen intention, record the smallest plausible root cause and select a revision that does not introduce a new obligation. If no participant evidence exists, describe the issue as a predicted risk or editorial hypothesis rather than an observed reader failure.

Failure 6: Actor swap

Detect: Does the reader assign approval or responsibility to the wrong party? Resolve: Draw an actor map based on the reader’s own explanation. The analyst should compare the actual evidence with the frozen intention, record the smallest plausible root cause and select a revision that does not introduce a new obligation. If no participant evidence exists, describe the issue as a predicted risk or editorial hypothesis rather than an observed reader failure.

Failure 7: Negation reversal

Detect: Does the target text invert permitted and prohibited actions? Resolve: Offer one allowed and one disallowed action and check both. The analyst should compare the actual evidence with the frozen intention, record the smallest plausible root cause and select a revision that does not introduce a new obligation. If no participant evidence exists, describe the issue as a predicted risk or editorial hypothesis rather than an observed reader failure.

Failure 8: Temporal collapse

Detect: Does the reader confuse submission, processing and effective dates? Resolve: Use three status snapshots across the sequence. The analyst should compare the actual evidence with the frozen intention, record the smallest plausible root cause and select a revision that does not introduce a new obligation. If no participant evidence exists, describe the issue as a predicted risk or editorial hypothesis rather than an observed reader failure.

Failure 9: False certainty

Detect: Does an estimate sound like a guarantee or a provisional state sound final? Resolve: Use an outcome in which the expectation does not materialize. The analyst should compare the actual evidence with the frozen intention, record the smallest plausible root cause and select a revision that does not introduce a new obligation. If no participant evidence exists, describe the issue as a predicted risk or editorial hypothesis rather than an observed reader failure.

Failure 10: Missing exception

Detect: Does the reader apply a rule when the exclusion condition is present? Resolve: Test a contrastive boundary pair differing only in the exception. The analyst should compare the actual evidence with the frozen intention, record the smallest plausible root cause and select a revision that does not introduce a new obligation. If no participant evidence exists, describe the issue as a predicted risk or editorial hypothesis rather than an observed reader failure.

Failure 11: Numerical anchoring

Detect: Are units, quantities or counting rules inferred differently? Resolve: Test just below, at and above the named threshold. The analyst should compare the actual evidence with the frozen intention, record the smallest plausible root cause and select a revision that does not introduce a new obligation. If no participant evidence exists, describe the issue as a predicted risk or editorial hypothesis rather than an observed reader failure.

Failure 12: Icon-text mismatch

Detect: Does an icon imply success while text describes a pending stage? Resolve: Use a task trace to identify which visual cue overrode the sentence. The analyst should compare the actual evidence with the frozen intention, record the smallest plausible root cause and select a revision that does not introduce a new obligation. If no participant evidence exists, describe the issue as a predicted risk or editorial hypothesis rather than an observed reader failure.

Failure 13: Survey construct drift

Detect: Does a response scale measure intensity when it should measure frequency? Resolve: Ask respondents what quantity their answer describes. The analyst should compare the actual evidence with the frozen intention, record the smallest plausible root cause and select a revision that does not introduce a new obligation. If no participant evidence exists, describe the issue as a predicted risk or editorial hypothesis rather than an observed reader failure.

Failure 14: Social desirability

Detect: Does the participant agree with the interviewer rather than demonstrate comprehension? Resolve: Remove approval cues and request a concrete independent choice. The analyst should compare the actual evidence with the frozen intention, record the smallest plausible root cause and select a revision that does not introduce a new obligation. If no participant evidence exists, describe the issue as a predicted risk or editorial hypothesis rather than an observed reader failure.

Failure 15: Expert overfitting

Detect: Do only professionals recognize a shorthand that general readers need to act on? Resolve: Recruit novices and specialists, then compare inferred rules. The analyst should compare the actual evidence with the frozen intention, record the smallest plausible root cause and select a revision that does not introduce a new obligation. If no participant evidence exists, describe the issue as a predicted risk or editorial hypothesis rather than an observed reader failure.

Failure 16: False security assurance

Detect: Does a privacy or security verb imply more protection than the product delivers? Resolve: Ask what remains accessible or processed after the action. The analyst should compare the actual evidence with the frozen intention, record the smallest plausible root cause and select a revision that does not introduce a new obligation. If no participant evidence exists, describe the issue as a predicted risk or editorial hypothesis rather than an observed reader failure.

Failure 17: Cross-language answer smoothing

Detect: Did the analyst’s translation erase hesitation, uncertainty or a double meaning? Resolve: Preserve original-language quotes and independently review pivotal evidence. The analyst should compare the actual evidence with the frozen intention, record the smallest plausible root cause and select a revision that does not introduce a new obligation. If no participant evidence exists, describe the issue as a predicted risk or editorial hypothesis rather than an observed reader failure.

Failure 18: Unretested repair

Detect: Was the revised target copy accepted only by editorial consensus? Resolve: Run the revised artifact on an independent realistic scenario. The analyst should compare the actual evidence with the frozen intention, record the smallest plausible root cause and select a revision that does not introduce a new obligation. If no participant evidence exists, describe the issue as a predicted risk or editorial hypothesis rather than an observed reader failure.

Failure 19: Too-small study claims

Detect: Has a handful of qualitative interviews been presented as statistical proof? Resolve: State sampling limits and use additional evidence proportional to risk. The analyst should compare the actual evidence with the frozen intention, record the smallest plausible root cause and select a revision that does not introduce a new obligation. If no participant evidence exists, describe the issue as a predicted risk or editorial hypothesis rather than an observed reader failure.

Failure 20: Mixed versions

Detect: Were participants shown different source rules or interface states unknowingly? Resolve: Version both content and screenshots in every observation. The analyst should compare the actual evidence with the frozen intention, record the smallest plausible root cause and select a revision that does not introduce a new obligation. If no participant evidence exists, describe the issue as a predicted risk or editorial hypothesis rather than an observed reader failure.

Failure 21: Synthetic cases mistaken for research

Detect: Were illustrative fictional examples presented as observed user behavior? Resolve: Label teaching fixtures and real observations separately. The analyst should compare the actual evidence with the frozen intention, record the smallest plausible root cause and select a revision that does not introduce a new obligation. If no participant evidence exists, describe the issue as a predicted risk or editorial hypothesis rather than an observed reader failure.

Failure 22: Over-broad remedy

Detect: Did the revision make one sentence easier but break a linked exception elsewhere? Resolve: Inspect dependencies and retest the affected workflow. The analyst should compare the actual evidence with the frozen intention, record the smallest plausible root cause and select a revision that does not introduce a new obligation. If no participant evidence exists, describe the issue as a predicted risk or editorial hypothesis rather than an observed reader failure.

Failure 23: No accountable owner

Detect: Is a linguist being asked to decide policy or safety intent? Resolve: Escalate the underlying rule to the responsible subject expert. The analyst should compare the actual evidence with the frozen intention, record the smallest plausible root cause and select a revision that does not introduce a new obligation. If no participant evidence exists, describe the issue as a predicted risk or editorial hypothesis rather than an observed reader failure.

Failure 24: All-or-nothing comprehension

Detect: Did the report count a correct click as full understanding despite a wrong explanation? Resolve: Separate action, rationale, expectation and confidence. The analyst should compare the actual evidence with the frozen intention, record the smallest plausible root cause and select a revision that does not introduce a new obligation. If no participant evidence exists, describe the issue as a predicted risk or editorial hypothesis rather than an observed reader failure.

Failure 25: Translation-by-preference

Detect: Was the most liked phrase selected despite inducing a wrong action? Resolve: Prioritize faithful meaning and task evidence over aesthetic votes. The analyst should compare the actual evidence with the frozen intention, record the smallest plausible root cause and select a revision that does not introduce a new obligation. If no participant evidence exists, describe the issue as a predicted risk or editorial hypothesis rather than an observed reader failure.

Failure 26: Accessibility conflation

Detect: Does a screen reader or focus-order defect get described as ‘bad translation’? Resolve: Test language realization and accessibility separately, then together. The analyst should compare the actual evidence with the frozen intention, record the smallest plausible root cause and select a revision that does not introduce a new obligation. If no participant evidence exists, describe the issue as a predicted risk or editorial hypothesis rather than an observed reader failure.

Failure 27: Retest contamination

Detect: Did previously coached respondents make the corrected version appear better? Resolve: Use fresh participants or document learning effects explicitly. The analyst should compare the actual evidence with the frozen intention, record the smallest plausible root cause and select a revision that does not introduce a new obligation. If no participant evidence exists, describe the issue as a predicted risk or editorial hypothesis rather than an observed reader failure.

Failure 28: Scope creep into another owner

Detect: Did a comprehension section begin replacing a vocabulary dictionary or an English grammar curriculum? Resolve: Explain only the terms required to complete the translation task and route outward. The analyst should compare the actual evidence with the frozen intention, record the smallest plausible root cause and select a revision that does not introduce a new obligation. If no participant evidence exists, describe the issue as a predicted risk or editorial hypothesis rather than an observed reader failure.

Part VI. Transfer: how to use the system outside this chapter

For a short interface label, begin with two realistic choices and one contrastive question. For a public notice, emphasize actor, deadline, eligibility and consequence. For a research instrument, preserve the measured construct and follow the appropriate specialist validation protocol. For instructions, test ordering and the points where a reader could reasonably stop or branch. For a document containing terms of art, keep the required precision while making any essential explanation visible. In every case, move from text similarity toward demonstrated understanding without mistaking qualitative evidence for universal certification.

A minimum result record contains: artifact ID and source revision; target language and version; exact message and surrounding context; intended interpretation contract; participant recruitment criteria; neutral script; observed paraphrases; actual task selections; boundary-test results; error attribution; revision rationale; qualified reviewer approval where required; and retest limitations. Store this with controlled access and respect participant confidentiality. Do not turn session recordings or personal details into public examples without lawful authorization.

This chapter is a practical system, not an official translation accreditation or a replacement for professional judgment. It does not establish universal sample sizes, guarantee comprehension in every language variety, or provide legal, medical or safety approval. Its useful promise is narrower and stronger: it teaches editors to ask questions for which a real reader can provide testable evidence.

Sources and next routes

• ISO 24495-1:2023 — Plain language: governing principles and guidelines (published June 2023; official standards summary).
• CDC Clear Communication Index — How to Use the Index (official guidance reviewed on 3 November 2026).
• CDC Clear Communication Index — FAQ and scoring scope (official guidance; Index scoring is distinct from reader testing).
• ISPOR Task Force (2005) — Translation and cultural adaptation of PRO measures (discipline-specific historical principles; not a universal sample-size rule).
• IEC/IEEE 82079-1:2019 — Preparation of information for use (official standard description).
• W3C — Understanding WCAG 2.2 (accessible language, labels and instructions).

For the broader architecture, return to Master Art of Translation | The Complete System for Moving Meaning Between Languages. For what happens after people report translation failures, follow the preceding chapter on multilingual translation error reporting and correction. The present chapter focuses on planned, pre-release evidence from reader interpretation rather than a generic post-release feedback queue.

If a failed test arises because a reader does not know the concept behind a word, explore the protected Vocabulary Learning Hub for concept development instead of turning this translation chapter into a dictionary. If the issue is specifically English grammar, inference or reading mechanics, use the protected How English Works learning map without taking ownership of those lessons. Both routes are supporting pathways; the domain of this chapter is multilingual reader-understanding validation.

Final principle: A translation is not validated merely because it can be read. It is better validated when representative readers, encountering an authentic task without coaching, reconstruct the intended rule and act on it for the correct reason—and when the limits of that evidence are stated plainly.