How do you translate Formative assessments without changing the meaning of questions or the fairness of their scores? Teachers working with multilingual Formative activities need to distinguish interface-language settings from translation of teacher-authored questions, answer choices, passages and feedback. If a translated question changes a negative, accepted-answer rule or relationship between two parts, students can receive different scores despite demonstrating the same understanding. That makes assessment translation an educational decision, not merely a language preference.
Formative allows teachers and students to choose an interface language, but its documented language setting applies to native buttons and menus rather than to text entered by a teacher. Its Student Translator is a separate feature associated with specified licensing. Schools should check their actual access before relying on it. The careful approach is to localize question banks, scoring criteria, rubrics and feedback together, then test the student experience under the intended conditions of assistance and pacing.
This detailed guide is written for a teacher or assessment coordinator making Formative activities intelligible while preserving the skill and standard the activity is intended to measure. It explains how to inspect an activity, preserve logical meaning, plan accessibility, verify learner-facing settings and interpret results with care. It includes concrete classroom failures, a worked teaching example, a four-week plan, questions often asked by teachers and official product references. The aim is a clear and useful multilingual learning experience, not a collection of disconnected translated strings.
The central translation boundary in Formative
Formative’s account language preference changes native interface controls, not teacher-authored questions. The Student Translator is a separate tool described in official documentation as dependent on certain Balanced Assessment licensing; availability should be checked in the actual school account.
In this environment, localization touches several kinds of material: Formative language settings, Student Translator, multi-part assessments, question keys, scoring, grading rubrics and pacing. Some of those elements belong to the software, some are under the teacher’s editorial control, and others are contributed by external publishers or learners. If a feature changes only one group, the rest must be handled intentionally. A reliable teacher does not promise a capability merely because a similarly named feature exists on another platform.
What does successful localization actually mean?
Successful educational localization makes the intended knowledge reachable through a language and interaction that the learner can understand. It does not require every sentence to preserve English word order. It requires instructions to remain accurate, sources to remain traceable and any resulting score or contribution to mean what the teacher says it means.
A class may legitimately receive different surface wording, reading support or context examples. But those differences must be chosen to help learners reach an equivalent objective rather than inadvertently changing the objective. Begin by writing down the concept and the action expected from the student. Without that anchor, even excellent translation can drift.
Map the complete learning journey
Begin with the first link the learner receives. Follow the title, introduction, task material, media, any external evidence, response interface, feedback and final display or report. Identify what is written by a teacher, what belongs to the platform, what comes from an external publisher and what was contributed by a student. Each has a different editorial responsibility.
This simple inventory prevents a common disappointment: a beautiful translated introduction followed by an English-only resource or an inaccessible response step. It also helps identify when an automatic tool is translating an interface, when it is translating user content and when it does not reach the relevant material at all.
Preserve logical and factual relationships
Words such as not, only, usually, at least, before and after do a surprising amount of educational work. Translating them incorrectly can reverse a correct answer, change a scientific claim or make a historical sequence impossible. Preserve numerical values, units, names, source attributions and uncertainty, and ask whether the finished passage still supports the intended conclusion.
Terminology should suit the intended curriculum and the reader’s stage of learning. A formal technical term may be right in one subject and obscure in a beginner lesson. Prefer a stable term list over uncontrolled synonyms; where several valid variants exist, acknowledge them without pretending that all alternatives mean exactly the same thing.
Distinguish reading help from answer help
An assistance feature may be appropriate in a science task where the construct is reasoning, but inappropriate in a vocabulary quiz where recognizing an unaided word is the construct. Translate the directions that are incidental to the subject, while protecting the information the learner is supposed to retrieve or infer independently.
A fair policy is specific. State whether participants may use translation, text-to-speech, subtitles, dictionaries or collaboration. Apply the same principle to each language group. The goal is not to ban helpful technology; it is to know what technology contributes to the evidence of learning.
Make accessible alternatives meaningful
A picture may contain crucial labels; a recording may convey a named person or a negated claim; a video may depend on the sequence of its scenes. A blank alt-text label, unverified automatic caption or unsuitable speech voice does not preserve the learning experience. Write or source an alternative that communicates the relevant information without silently giving away the answer.
Check a realistic participant’s device. Some browser voices are unavailable; some embedded sources are blocked; some long translated buttons wrap awkwardly. The editor may appear correct even while a student using a keyboard or mobile screen cannot finish. A short independent test can reveal what visual proofreading misses.
Protect authorship, privacy and source rights
When students or outside authors contribute material, keep their original claims distinct from a translator’s interpretation and a teacher’s summary. A multilingual display can be useful without making automatic translation authoritative. Attribute sources and preserve uncertainty, particularly in historical, scientific and public discussions.
Sharing is also an educational decision. Confirm whether the intended audience can view, contribute or edit, and inspect what a logged-out visitor sees. Do not expose student names or identifying data merely because a public link is convenient. Respect external publication rights rather than copying full sources without permission.
Review learning evidence rather than just completion
A dashboard score, count of posts or number of finished items may be useful, but it is not a complete explanation of learning. Compare the question, permitted assistance, source materials and scoring across languages before judging performances comparable. A translated option that is easier than the source can inflate marks; an untranslated source can suppress them.
Study representative mistakes and ask students to describe their understanding. When confusion comes from translation rather than subject knowledge, correct the task and avoid drawing a strong conclusion from an unreliable result. Good evidence allows teachers to explain not only what happened but why the result deserves trust.
Make improvements controlled and observable
Create clearly named editions by language and audience, then make a small set of purposeful changes. Check the entire journey after any meaningful edit, including sources and response rules that depend on the revised text. A version label is useful only if the editor knows what it represents.
Review comments from actual participants after release. Repeated questions about one term or one inaccessible step often identify a better improvement than a broad rewrite. This disciplined cycle makes localization sustainable for a small teaching team and helps avoid introducing new inconsistencies elsewhere.
Twenty-six practical Formative translation problems and solutions
The following cases are written as practical decisions, not as a claim that every classroom uses every feature. Each begins with a recognizable risk, explains how to respond and describes how to verify the result in an actual learner experience. Use the cases selectively: a carefully checked solution to the problem you have is worth more than a long untested list of possible features.
1. Interface language changes but questions do not
The practical problem is this: Teachers think changing account language will translate their authored activities. Translation quality is measured by whether the learner can perform the intended intellectual task, not by the number of words that appear in the chosen language. Ask what evidence the activity is meant to elicit and whether the learner has access to the same essential information. A fluent sentence may still be educationally wrong if it changes the response expected. The question is not simply whether a phrase looks right. It is whether a participant can work with the intended information and reach a response that means the same thing educationally.
The useful correction is to check interface controls and educational content separately and edit the actual assessment. Build one complete participant journey, from opening the shared link to receiving feedback or seeing a published contribution. If the proposed fix works only in an editor or in the owner’s privileged account, it is not yet a dependable learning solution. The actual audience view is decisive. The change should be explained in terms of what a learner is now able to understand or do, not in terms of the number of strings that were replaced.
Test the experience using a fresh or logged-out session as appropriate to the planned sharing model. Check the beginning, the action required and the final feedback or displayed contribution. A teacher account often hides access restrictions or special buttons that an ordinary student will encounter. In this particular situation, test the repaired interface language changes but questions do not condition explicitly rather than accepting a generic preview as evidence. If a question, source or control still changes the task, revise the learner-facing edition before it is used for grading or public interpretation.
2. Student Translator is not licensed
The practical problem is this: A lesson assumes every student sees a translator button. A teacher should distinguish the language of the instruction from the knowledge the question is intended to measure. Sometimes words are the learning objective; sometimes they are only the route to a mathematical, scientific or historical idea. That distinction determines whether a translation aid is fair or whether it gives away the answer. The question is not simply whether a phrase looks right. It is whether a participant can work with the intended information and reach a response that means the same thing educationally.
The useful correction is to verify available features and provide an approved reviewed-language path. Protect the relationship between the instruction, the evidence available and the response expected. Translating words literally may be less faithful than a careful rewording that preserves these relationships. When an example or interaction must change, explain the educational reason in the teaching team’s change record. The change should be explained in terms of what a learner is now able to understand or do, not in terms of the number of strings that were replaced.
Compare the original and translated learning evidence rather than relying on a score or word count alone. What can the learner infer, explain or do? Which supports were available? If those conditions differ materially, explain the limit on comparison instead of treating numbers as interchangeable. In this particular situation, test the repaired student translator is not licensed condition explicitly rather than accepting a generic preview as evidence. If a question, source or control still changes the task, revise the learner-facing edition before it is used for grading or public interpretation.
3. A superlative changes direction
The practical problem is this: The phrase least likely is rendered as most likely. The most revealing mistake is often small: a missing negative, a changed unit, an implied promise, a confusing button, an uncredited quotation. Such details may escape a quick proofreading pass because the passage sounds natural. Review meaning and consequence together before judging the edition ready. The question is not simply whether a phrase looks right. It is whether a participant can work with the intended information and reach a response that means the same thing educationally.
The useful correction is to review logic words and independently solve the translated question. Consider a second kind of learner before approving the solution: someone using another device, someone reading through an accessibility aid or someone who has not seen the source language. Their experience may reveal hidden information that the author assumed everyone already knew. Make the essential path self-contained. The change should be explained in terms of what a learner is now able to understand or do, not in terms of the number of strings that were replaced.
To verify the outcome, ask a reviewer who did not write the activity to follow the final participant link without extra explanation. The reviewer should say what the task asks, identify the evidence used and describe the action that completes it. A mismatch with the teacher’s intended task is grounds for revision. In this particular situation, test the repaired a superlative changes direction condition explicitly rather than accepting a generic preview as evidence. If a question, source or control still changes the task, revise the learner-facing edition before it is used for grading or public interpretation.
4. A negative becomes positive
The practical problem is this: A true-or-false question changes meaning after translation. Language reviewers and subject experts answer different questions. One can check natural wording, register and cultural clarity, while another checks the learning objective, accepted answers and evidence. Neither check should substitute entirely for the other in a consequential educational activity. The question is not simply whether a phrase looks right. It is whether a participant can work with the intended information and reach a response that means the same thing educationally.
The useful correction is to reassess the proposition and correct its final truth value. Use a limited, testable repair rather than a large unverified rewrite. A short, clear explanation, a more accurate key, a suitable alternative source or a correctly configured permission can solve the real problem. Once the change is stable, review nearby items that rely on the same language assumption. The change should be explained in terms of what a learner is now able to understand or do, not in terms of the number of strings that were replaced.
Check more than the most obvious correct path. Try a plausible misunderstanding, a typical device and the assistance a real participant can use. A translation may appear fine until the second step reveals a scoring mistake, missing source or unexpected permission. Record the observed result before release. In this particular situation, test the repaired a negative becomes positive condition explicitly rather than accepting a generic preview as evidence. If a question, source or control still changes the task, revise the learner-facing edition before it is used for grading or public interpretation.
5. Answer choices are reordered
The practical problem is this: A key remains connected to a previous option position. The learner’s route includes more than the paragraph being translated. There may be media, accessibility controls, a response form, permissions, a link to another website and feedback after completion. An effective editor treats those parts as a connected experience, not as unrelated fragments. The question is not simply whether a phrase looks right. It is whether a participant can work with the intended information and reach a response that means the same thing educationally.
The useful correction is to check each choice and its scoring after language editing. Start by writing a single sentence describing what learners should understand or demonstrate. Identify which words, media or settings control that outcome. Change the relevant element, and compare the revised activity against the original purpose rather than trusting linguistic similarity alone. Keep a source edition for careful cross-checking. The change should be explained in terms of what a learner is now able to understand or do, not in terms of the number of strings that were replaced.
Use a practical acceptance test that can be repeated after an edit: the language and version tested, the learner role, the expected response and the observed response. That small record is enough to distinguish a verified correction from a confident assumption in later classroom use. In this particular situation, test the repaired answer choices are reordered condition explicitly rather than accepting a generic preview as evidence. If a question, source or control still changes the task, revise the learner-facing edition before it is used for grading or public interpretation.
6. A case-sensitive answer is unfair
The practical problem is this: Concept knowledge is assessed by capitalization accidentally. The most dependable projects make their limits visible. If an automatic feature affects only part of the screen, say so. If a source cannot be accessed, replace it lawfully or explain the gap. If language editions have different assistance conditions, record that difference before interpreting outcomes. The question is not simply whether a phrase looks right. It is whether a participant can work with the intended information and reach a response that means the same thing educationally.
The useful correction is to apply case sensitivity only when it is part of the intended skill. Make the decision as a teacher and translator together. First inspect the source and the destination wording. Then identify whether the challenge is terminology, navigation, response rules, source access or cultural context. Address that root problem rather than adding more text to hide a confusing design. The change should be explained in terms of what a learner is now able to understand or do, not in terms of the number of strings that were replaced.
Ask both a language-aware reader and a relevant subject reviewer to explain the resulting behavior. Their judgments may differ for a useful reason: a sentence can be idiomatic yet misstate the subject. Resolve that difference before exposing students to a consequential activity. In this particular situation, test the repaired a case-sensitive answer is unfair condition explicitly rather than accepting a generic preview as evidence. If a question, source or control still changes the task, revise the learner-facing edition before it is used for grading or public interpretation.
7. Grammar variants get rejected
The practical problem is this: The accepted answer includes a single inflected surface form. Translation quality is measured by whether the learner can perform the intended intellectual task, not by the number of words that appear in the chosen language. Ask what evidence the activity is meant to elicit and whether the learner has access to the same essential information. A fluent sentence may still be educationally wrong if it changes the response expected. The question is not simply whether a phrase looks right. It is whether a participant can work with the intended information and reach a response that means the same thing educationally.
The useful correction is to test legitimate variants or review the answer manually. Build one complete participant journey, from opening the shared link to receiving feedback or seeing a published contribution. If the proposed fix works only in an editor or in the owner’s privileged account, it is not yet a dependable learning solution. The actual audience view is decisive. The change should be explained in terms of what a learner is now able to understand or do, not in terms of the number of strings that were replaced.
Test the experience using a fresh or logged-out session as appropriate to the planned sharing model. Check the beginning, the action required and the final feedback or displayed contribution. A teacher account often hides access restrictions or special buttons that an ordinary student will encounter. In this particular situation, test the repaired grammar variants get rejected condition explicitly rather than accepting a generic preview as evidence. If a question, source or control still changes the task, revise the learner-facing edition before it is used for grading or public interpretation.
8. A decimal is entered differently
The practical problem is this: A correct quantity uses the target locale’s separator. A teacher should distinguish the language of the instruction from the knowledge the question is intended to measure. Sometimes words are the learning objective; sometimes they are only the route to a mathematical, scientific or historical idea. That distinction determines whether a translation aid is fair or whether it gives away the answer. The question is not simply whether a phrase looks right. It is whether a participant can work with the intended information and reach a response that means the same thing educationally.
The useful correction is to make notation explicit and check numeric grading behavior. Protect the relationship between the instruction, the evidence available and the response expected. Translating words literally may be less faithful than a careful rewording that preserves these relationships. When an example or interaction must change, explain the educational reason in the teaching team’s change record. The change should be explained in terms of what a learner is now able to understand or do, not in terms of the number of strings that were replaced.
Compare the original and translated learning evidence rather than relying on a score or word count alone. What can the learner infer, explain or do? Which supports were available? If those conditions differ materially, explain the limit on comparison instead of treating numbers as interchangeable. In this particular situation, test the repaired a decimal is entered differently condition explicitly rather than accepting a generic preview as evidence. If a question, source or control still changes the task, revise the learner-facing edition before it is used for grading or public interpretation.
9. A question image uses English labels
The practical problem is this: The diagram is needed but text embedded in it remains untranslated. The most revealing mistake is often small: a missing negative, a changed unit, an implied promise, a confusing button, an uncredited quotation. Such details may escape a quick proofreading pass because the passage sounds natural. Review meaning and consequence together before judging the edition ready. The question is not simply whether a phrase looks right. It is whether a participant can work with the intended information and reach a response that means the same thing educationally.
The useful correction is to provide localized labels or equivalent accessible descriptions. Consider a second kind of learner before approving the solution: someone using another device, someone reading through an accessibility aid or someone who has not seen the source language. Their experience may reveal hidden information that the author assumed everyone already knew. Make the essential path self-contained. The change should be explained in terms of what a learner is now able to understand or do, not in terms of the number of strings that were replaced.
To verify the outcome, ask a reviewer who did not write the activity to follow the final participant link without extra explanation. The reviewer should say what the task asks, identify the evidence used and describe the action that completes it. A mismatch with the teacher’s intended task is grounds for revision. In this particular situation, test the repaired a question image uses english labels condition explicitly rather than accepting a generic preview as evidence. If a question, source or control still changes the task, revise the learner-facing edition before it is used for grading or public interpretation.
10. Part A and Part B become inconsistent
The practical problem is this: An evidence selection no longer supports the translated inference. Language reviewers and subject experts answer different questions. One can check natural wording, register and cultural clarity, while another checks the learning objective, accepted answers and evidence. Neither check should substitute entirely for the other in a consequential educational activity. The question is not simply whether a phrase looks right. It is whether a participant can work with the intended information and reach a response that means the same thing educationally.
The useful correction is to translate multi-part items and source stimulus together. Use a limited, testable repair rather than a large unverified rewrite. A short, clear explanation, a more accurate key, a suitable alternative source or a correctly configured permission can solve the real problem. Once the change is stable, review nearby items that rely on the same language assumption. The change should be explained in terms of what a learner is now able to understand or do, not in terms of the number of strings that were replaced.
Check more than the most obvious correct path. Try a plausible misunderstanding, a typical device and the assistance a real participant can use. A translation may appear fine until the second step reveals a scoring mistake, missing source or unexpected permission. Record the observed result before release. In this particular situation, test the repaired part a and part b become inconsistent condition explicitly rather than accepting a generic preview as evidence. If a question, source or control still changes the task, revise the learner-facing edition before it is used for grading or public interpretation.
11. A citation in the stimulus is lost
The practical problem is this: The learner cannot identify which evidence passage to use. The learner’s route includes more than the paragraph being translated. There may be media, accessibility controls, a response form, permissions, a link to another website and feedback after completion. An effective editor treats those parts as a connected experience, not as unrelated fragments. The question is not simply whether a phrase looks right. It is whether a participant can work with the intended information and reach a response that means the same thing educationally.
The useful correction is to retain source identity and make the reference clear. Start by writing a single sentence describing what learners should understand or demonstrate. Identify which words, media or settings control that outcome. Change the relevant element, and compare the revised activity against the original purpose rather than trusting linguistic similarity alone. Keep a source edition for careful cross-checking. The change should be explained in terms of what a learner is now able to understand or do, not in terms of the number of strings that were replaced.
Use a practical acceptance test that can be repeated after an edit: the language and version tested, the learner role, the expected response and the observed response. That small record is enough to distinguish a verified correction from a confident assumption in later classroom use. In this particular situation, test the repaired a citation in the stimulus is lost condition explicitly rather than accepting a generic preview as evidence. If a question, source or control still changes the task, revise the learner-facing edition before it is used for grading or public interpretation.
12. Randomization breaks reasoning order
The practical problem is this: Questions depend on information shown earlier. The most dependable projects make their limits visible. If an automatic feature affects only part of the screen, say so. If a source cannot be accessed, replace it lawfully or explain the gap. If language editions have different assistance conditions, record that difference before interpreting outcomes. The question is not simply whether a phrase looks right. It is whether a participant can work with the intended information and reach a response that means the same thing educationally.
The useful correction is to use the appropriate multi-part structure and check the learner sequence. Make the decision as a teacher and translator together. First inspect the source and the destination wording. Then identify whether the challenge is terminology, navigation, response rules, source access or cultural context. Address that root problem rather than adding more text to hide a confusing design. The change should be explained in terms of what a learner is now able to understand or do, not in terms of the number of strings that were replaced.
Ask both a language-aware reader and a relevant subject reviewer to explain the resulting behavior. Their judgments may differ for a useful reason: a sentence can be idiomatic yet misstate the subject. Resolve that difference before exposing students to a consequential activity. In this particular situation, test the repaired randomization breaks reasoning order condition explicitly rather than accepting a generic preview as evidence. If a question, source or control still changes the task, revise the learner-facing edition before it is used for grading or public interpretation.
13. Teacher-paced reading is too fast
The practical problem is this: A class advances before target-language learners can read the prompt. Translation quality is measured by whether the learner can perform the intended intellectual task, not by the number of words that appear in the chosen language. Ask what evidence the activity is meant to elicit and whether the learner has access to the same essential information. A fluent sentence may still be educationally wrong if it changes the response expected. The question is not simply whether a phrase looks right. It is whether a participant can work with the intended information and reach a response that means the same thing educationally.
The useful correction is to pilot timing and adjust pace without changing construct. Build one complete participant journey, from opening the shared link to receiving feedback or seeing a published contribution. If the proposed fix works only in an editor or in the owner’s privileged account, it is not yet a dependable learning solution. The actual audience view is decisive. The change should be explained in terms of what a learner is now able to understand or do, not in terms of the number of strings that were replaced.
Test the experience using a fresh or logged-out session as appropriate to the planned sharing model. Check the beginning, the action required and the final feedback or displayed contribution. A teacher account often hides access restrictions or special buttons that an ordinary student will encounter. In this particular situation, test the repaired teacher-paced reading is too fast condition explicitly rather than accepting a generic preview as evidence. If a question, source or control still changes the task, revise the learner-facing edition before it is used for grading or public interpretation.
14. Student-paced checkpoints are unclear
The practical problem is this: Learners cannot tell when to submit or review feedback. A teacher should distinguish the language of the instruction from the knowledge the question is intended to measure. Sometimes words are the learning objective; sometimes they are only the route to a mathematical, scientific or historical idea. That distinction determines whether a translation aid is fair or whether it gives away the answer. The question is not simply whether a phrase looks right. It is whether a participant can work with the intended information and reach a response that means the same thing educationally.
The useful correction is to provide concise accessible instructions for each step. Protect the relationship between the instruction, the evidence available and the response expected. Translating words literally may be less faithful than a careful rewording that preserves these relationships. When an example or interaction must change, explain the educational reason in the teaching team’s change record. The change should be explained in terms of what a learner is now able to understand or do, not in terms of the number of strings that were replaced.
Compare the original and translated learning evidence rather than relying on a score or word count alone. What can the learner infer, explain or do? Which supports were available? If those conditions differ materially, explain the limit on comparison instead of treating numbers as interchangeable. In this particular situation, test the repaired student-paced checkpoints are unclear condition explicitly rather than accepting a generic preview as evidence. If a question, source or control still changes the task, revise the learner-facing edition before it is used for grading or public interpretation.
15. Audio response changes the target skill
The practical problem is this: An oral explanation task becomes reading an external script. The most revealing mistake is often small: a missing negative, a changed unit, an implied promise, a confusing button, an uncredited quotation. Such details may escape a quick proofreading pass because the passage sounds natural. Review meaning and consequence together before judging the edition ready. The question is not simply whether a phrase looks right. It is whether a participant can work with the intended information and reach a response that means the same thing educationally.
The useful correction is to define whether spontaneity, reasoning or pronunciation is assessed. Consider a second kind of learner before approving the solution: someone using another device, someone reading through an accessibility aid or someone who has not seen the source language. Their experience may reveal hidden information that the author assumed everyone already knew. Make the essential path self-contained. The change should be explained in terms of what a learner is now able to understand or do, not in terms of the number of strings that were replaced.
To verify the outcome, ask a reviewer who did not write the activity to follow the final participant link without extra explanation. The reviewer should say what the task asks, identify the evidence used and describe the action that completes it. A mismatch with the teacher’s intended task is grounds for revision. In this particular situation, test the repaired audio response changes the target skill condition explicitly rather than accepting a generic preview as evidence. If a question, source or control still changes the task, revise the learner-facing edition before it is used for grading or public interpretation.
16. The passage is too technical
The practical problem is this: Translation introduces vocabulary above the target grade. Language reviewers and subject experts answer different questions. One can check natural wording, register and cultural clarity, while another checks the learning objective, accepted answers and evidence. Neither check should substitute entirely for the other in a consequential educational activity. The question is not simply whether a phrase looks right. It is whether a participant can work with the intended information and reach a response that means the same thing educationally.
The useful correction is to match the learning level and curriculum terminology. Use a limited, testable repair rather than a large unverified rewrite. A short, clear explanation, a more accurate key, a suitable alternative source or a correctly configured permission can solve the real problem. Once the change is stable, review nearby items that rely on the same language assumption. The change should be explained in terms of what a learner is now able to understand or do, not in terms of the number of strings that were replaced.
Check more than the most obvious correct path. Try a plausible misunderstanding, a typical device and the assistance a real participant can use. A translation may appear fine until the second step reveals a scoring mistake, missing source or unexpected permission. Record the observed result before release. In this particular situation, test the repaired the passage is too technical condition explicitly rather than accepting a generic preview as evidence. If a question, source or control still changes the task, revise the learner-facing edition before it is used for grading or public interpretation.
17. Rubric adjectives are vague
The practical problem is this: Translations of developing and proficient produce different expectations. The learner’s route includes more than the paragraph being translated. There may be media, accessibility controls, a response form, permissions, a link to another website and feedback after completion. An effective editor treats those parts as a connected experience, not as unrelated fragments. The question is not simply whether a phrase looks right. It is whether a participant can work with the intended information and reach a response that means the same thing educationally.
The useful correction is to write observable performance descriptions and moderate responses. Start by writing a single sentence describing what learners should understand or demonstrate. Identify which words, media or settings control that outcome. Change the relevant element, and compare the revised activity against the original purpose rather than trusting linguistic similarity alone. Keep a source edition for careful cross-checking. The change should be explained in terms of what a learner is now able to understand or do, not in terms of the number of strings that were replaced.
Use a practical acceptance test that can be repeated after an edit: the language and version tested, the learner role, the expected response and the observed response. That small record is enough to distinguish a verified correction from a confident assumption in later classroom use. In this particular situation, test the repaired rubric adjectives are vague condition explicitly rather than accepting a generic preview as evidence. If a question, source or control still changes the task, revise the learner-facing edition before it is used for grading or public interpretation.
18. Feedback describes an old distractor
The practical problem is this: An answer choice changed but the explanation did not. The most dependable projects make their limits visible. If an automatic feature affects only part of the screen, say so. If a source cannot be accessed, replace it lawfully or explain the gap. If language editions have different assistance conditions, record that difference before interpreting outcomes. The question is not simply whether a phrase looks right. It is whether a participant can work with the intended information and reach a response that means the same thing educationally.
The useful correction is to run incorrect and correct test answers through feedback. Make the decision as a teacher and translator together. First inspect the source and the destination wording. Then identify whether the challenge is terminology, navigation, response rules, source access or cultural context. Address that root problem rather than adding more text to hide a confusing design. The change should be explained in terms of what a learner is now able to understand or do, not in terms of the number of strings that were replaced.
Ask both a language-aware reader and a relevant subject reviewer to explain the resulting behavior. Their judgments may differ for a useful reason: a sentence can be idiomatic yet misstate the subject. Resolve that difference before exposing students to a consequential activity. In this particular situation, test the repaired feedback describes an old distractor condition explicitly rather than accepting a generic preview as evidence. If a question, source or control still changes the task, revise the learner-facing edition before it is used for grading or public interpretation.
19. Hints become answers
The practical problem is this: A translated hint states the solution explicitly. Translation quality is measured by whether the learner can perform the intended intellectual task, not by the number of words that appear in the chosen language. Ask what evidence the activity is meant to elicit and whether the learner has access to the same essential information. A fluent sentence may still be educationally wrong if it changes the response expected. The question is not simply whether a phrase looks right. It is whether a participant can work with the intended information and reach a response that means the same thing educationally.
The useful correction is to review help strength and decide whether it is permitted. Build one complete participant journey, from opening the shared link to receiving feedback or seeing a published contribution. If the proposed fix works only in an editor or in the owner’s privileged account, it is not yet a dependable learning solution. The actual audience view is decisive. The change should be explained in terms of what a learner is now able to understand or do, not in terms of the number of strings that were replaced.
Test the experience using a fresh or logged-out session as appropriate to the planned sharing model. Check the beginning, the action required and the final feedback or displayed contribution. A teacher account often hides access restrictions or special buttons that an ordinary student will encounter. In this particular situation, test the repaired hints become answers condition explicitly rather than accepting a generic preview as evidence. If a question, source or control still changes the task, revise the learner-facing edition before it is used for grading or public interpretation.
20. An AI-generated question changes curriculum scope
The practical problem is this: A fluent new question tests a different objective. A teacher should distinguish the language of the instruction from the knowledge the question is intended to measure. Sometimes words are the learning objective; sometimes they are only the route to a mathematical, scientific or historical idea. That distinction determines whether a translation aid is fair or whether it gives away the answer. The question is not simply whether a phrase looks right. It is whether a participant can work with the intended information and reach a response that means the same thing educationally.
The useful correction is to compare every generated item against approved evidence and standards. Protect the relationship between the instruction, the evidence available and the response expected. Translating words literally may be less faithful than a careful rewording that preserves these relationships. When an example or interaction must change, explain the educational reason in the teaching team’s change record. The change should be explained in terms of what a learner is now able to understand or do, not in terms of the number of strings that were replaced.
Compare the original and translated learning evidence rather than relying on a score or word count alone. What can the learner infer, explain or do? Which supports were available? If those conditions differ materially, explain the limit on comparison instead of treating numbers as interchangeable. In this particular situation, test the repaired an ai-generated question changes curriculum scope condition explicitly rather than accepting a generic preview as evidence. If a question, source or control still changes the task, revise the learner-facing edition before it is used for grading or public interpretation.
21. A learner uses a regional synonym
The practical problem is this: A valid answer is marked wrong due to narrow wording. The most revealing mistake is often small: a missing negative, a changed unit, an implied promise, a confusing button, an uncredited quotation. Such details may escape a quick proofreading pass because the passage sounds natural. Review meaning and consequence together before judging the edition ready. The question is not simply whether a phrase looks right. It is whether a participant can work with the intended information and reach a response that means the same thing educationally.
The useful correction is to include legitimate variants or grade for conceptual content. Consider a second kind of learner before approving the solution: someone using another device, someone reading through an accessibility aid or someone who has not seen the source language. Their experience may reveal hidden information that the author assumed everyone already knew. Make the essential path self-contained. The change should be explained in terms of what a learner is now able to understand or do, not in terms of the number of strings that were replaced.
To verify the outcome, ask a reviewer who did not write the activity to follow the final participant link without extra explanation. The reviewer should say what the task asks, identify the evidence used and describe the action that completes it. A mismatch with the teacher’s intended task is grounds for revision. In this particular situation, test the repaired a learner uses a regional synonym condition explicitly rather than accepting a generic preview as evidence. If a question, source or control still changes the task, revise the learner-facing edition before it is used for grading or public interpretation.
22. A score report overstates mastery
The practical problem is this: Parents mistake a quick practice activity for formal assessment. Language reviewers and subject experts answer different questions. One can check natural wording, register and cultural clarity, while another checks the learning objective, accepted answers and evidence. Neither check should substitute entirely for the other in a consequential educational activity. The question is not simply whether a phrase looks right. It is whether a participant can work with the intended information and reach a response that means the same thing educationally.
The useful correction is to explain what the particular activity can support as evidence. Use a limited, testable repair rather than a large unverified rewrite. A short, clear explanation, a more accurate key, a suitable alternative source or a correctly configured permission can solve the real problem. Once the change is stable, review nearby items that rely on the same language assumption. The change should be explained in terms of what a learner is now able to understand or do, not in terms of the number of strings that were replaced.
Check more than the most obvious correct path. Try a plausible misunderstanding, a typical device and the assistance a real participant can use. A translation may appear fine until the second step reveals a scoring mistake, missing source or unexpected permission. Record the observed result before release. In this particular situation, test the repaired a score report overstates mastery condition explicitly rather than accepting a generic preview as evidence. If a question, source or control still changes the task, revise the learner-facing edition before it is used for grading or public interpretation.
23. A response gallery reveals identity
The practical problem is this: Shared student answers show names unnecessarily. The learner’s route includes more than the paragraph being translated. There may be media, accessibility controls, a response form, permissions, a link to another website and feedback after completion. An effective editor treats those parts as a connected experience, not as unrelated fragments. The question is not simply whether a phrase looks right. It is whether a participant can work with the intended information and reach a response that means the same thing educationally.
The useful correction is to review privacy before presenting responses to the class. Start by writing a single sentence describing what learners should understand or demonstrate. Identify which words, media or settings control that outcome. Change the relevant element, and compare the revised activity against the original purpose rather than trusting linguistic similarity alone. Keep a source edition for careful cross-checking. The change should be explained in terms of what a learner is now able to understand or do, not in terms of the number of strings that were replaced.
Use a practical acceptance test that can be repeated after an edit: the language and version tested, the learner role, the expected response and the observed response. That small record is enough to distinguish a verified correction from a confident assumption in later classroom use. In this particular situation, test the repaired a response gallery reveals identity condition explicitly rather than accepting a generic preview as evidence. If a question, source or control still changes the task, revise the learner-facing edition before it is used for grading or public interpretation.
24. One language edition is easier
The practical problem is this: A target version includes an extra example or clue. The most dependable projects make their limits visible. If an automatic feature affects only part of the screen, say so. If a source cannot be accessed, replace it lawfully or explain the gap. If language editions have different assistance conditions, record that difference before interpreting outcomes. The question is not simply whether a phrase looks right. It is whether a participant can work with the intended information and reach a response that means the same thing educationally.
The useful correction is to record changes and check cognitive difficulty side by side. Make the decision as a teacher and translator together. First inspect the source and the destination wording. Then identify whether the challenge is terminology, navigation, response rules, source access or cultural context. Address that root problem rather than adding more text to hide a confusing design. The change should be explained in terms of what a learner is now able to understand or do, not in terms of the number of strings that were replaced.
Ask both a language-aware reader and a relevant subject reviewer to explain the resulting behavior. Their judgments may differ for a useful reason: a sentence can be idiomatic yet misstate the subject. Resolve that difference before exposing students to a consequential activity. In this particular situation, test the repaired one language edition is easier condition explicitly rather than accepting a generic preview as evidence. If a question, source or control still changes the task, revise the learner-facing edition before it is used for grading or public interpretation.
25. The interface uses another dialect
The practical problem is this: Navigation and question instructions use inconsistent variants. Translation quality is measured by whether the learner can perform the intended intellectual task, not by the number of words that appear in the chosen language. Ask what evidence the activity is meant to elicit and whether the learner has access to the same essential information. A fluent sentence may still be educationally wrong if it changes the response expected. The question is not simply whether a phrase looks right. It is whether a participant can work with the intended information and reach a response that means the same thing educationally.
The useful correction is to prioritize clarity and use a stable terminology list. Build one complete participant journey, from opening the shared link to receiving feedback or seeing a published contribution. If the proposed fix works only in an editor or in the owner’s privileged account, it is not yet a dependable learning solution. The actual audience view is decisive. The change should be explained in terms of what a learner is now able to understand or do, not in terms of the number of strings that were replaced.
Test the experience using a fresh or logged-out session as appropriate to the planned sharing model. Check the beginning, the action required and the final feedback or displayed contribution. A teacher account often hides access restrictions or special buttons that an ordinary student will encounter. In this particular situation, test the repaired the interface uses another dialect condition explicitly rather than accepting a generic preview as evidence. If a question, source or control still changes the task, revise the learner-facing edition before it is used for grading or public interpretation.
26. Different assistance conditions distort comparison
The practical problem is this: One group uses a translator and another has no comparable support. A teacher should distinguish the language of the instruction from the knowledge the question is intended to measure. Sometimes words are the learning objective; sometimes they are only the route to a mathematical, scientific or historical idea. That distinction determines whether a translation aid is fair or whether it gives away the answer. The question is not simply whether a phrase looks right. It is whether a participant can work with the intended information and reach a response that means the same thing educationally.
The useful correction is to interpret scores with known conditions and avoid unsupported rankings. Protect the relationship between the instruction, the evidence available and the response expected. Translating words literally may be less faithful than a careful rewording that preserves these relationships. When an example or interaction must change, explain the educational reason in the teaching team’s change record. The change should be explained in terms of what a learner is now able to understand or do, not in terms of the number of strings that were replaced.
Compare the original and translated learning evidence rather than relying on a score or word count alone. What can the learner infer, explain or do? Which supports were available? If those conditions differ materially, explain the limit on comparison instead of treating numbers as interchangeable. In this particular situation, test the repaired different assistance conditions distort comparison condition explicitly rather than accepting a generic preview as evidence. If a question, source or control still changes the task, revise the learner-facing edition before it is used for grading or public interpretation.
Worked example: a complete multilingual Formative lesson
A school adapts an ecosystems assessment for two language groups. Both groups should explain how a decline in one species affects a food web. The activity includes an evidence passage, multiple-choice items, a two-part explanation and teacher feedback. The assessment must distinguish science reasoning from the student’s speed of reading English.
Define the common lesson
Write the central learning objective in ordinary language. In the classroom example, the teacher specifies what each student should identify, justify or contribute, and what evidence would demonstrate success. The aim is not to translate every word first. It is to protect the intellectual work that will remain meaningful when the language changes.
Inspect the source version
Walk through the existing resource as a new participant. Note every title, instruction, linked source, audio cue, image label, response field and final result that shapes the task. Separate platform-generated text from teacher-written material and outside authors’ words. This inventory shows which language needs can actually be addressed by each available feature.
Prepare a checked target edition
Translate the teacher-controlled text with natural subject terminology, then inspect source access and any built-in language options. A subject expert checks logical claims and answer criteria while a language reader checks ease of comprehension. Where an example must change, record its purpose and verify that the student still encounters an equivalent challenge.
Pilot the complete learner journey
Ask a participant who has not seen the draft to open the delivered link and complete the task unaided. Test a typical response, a likely mistake and an accessible route. If the participant cannot tell what to do, the translation is not finished. If a result is recorded incorrectly, do not treat that result as trustworthy classroom evidence.
Improve without overstating results
Review the participant’s response and feedback against the intended objective. Was the necessary evidence accessible? Did a language support change the task? Did the platform produce the expected result? Correct the specific source of confusion in a new edition. Describe unresolved limitations plainly instead of treating a smooth interface as proof of perfect equivalence.
A four-week Formative localization plan
Week one — map and prioritize
Select one manageable lesson. Define the concept, the reader, the source language, the target language and the permitted help. Open every required resource and mark the moments where language changes behavior. The goal is a trustworthy inventory of the whole learner experience, including response rules and source permissions.
Choose several high-risk examples for testing: one negative or qualifier, one piece of subject vocabulary, one media item and one action at the end. Record the original purpose of each element. At the end of the week, the team should know exactly what needs adapting and what may not be changed without altering the learning objective.
Week two — localize and review
Create a target-language copy or version using the tool’s actual supported workflow. Translate teacher-controlled instructions, arrange accessible media and check terminology against the intended curriculum. Make no promises about features until they are visible in the school’s own account and device. Keep the original available when provenance or source comparison matters.
Ask a second reader to identify unclear wording and a subject teacher to inspect correct answers, evidence and difficulty. Recheck long labels, dates, names, units and permissions. If the localized material requires a changed example, state how it preserves the same concept. A complete tested short path matters more than many disconnected translated sentences.
Week three — pilot and correct
Give representative participants the actual link and let them proceed without spoken rescue. Watch where they hesitate, what they infer and whether they can reach the evidence needed to respond. Try accessibility features and at least one plausible incorrect path. Compare the system’s recorded result or published contribution with the task’s intention.
Separate language problems from subject misunderstandings before changing a question or interpreting a score. Test proposed corrections with a fresh participant where possible. Record what was changed and what remains uncertain. By the end of the week, the team should be able to explain the lesson’s strengths and limitations in ordinary language.
Week four — publish and learn
Check sharing rights, privacy, attribution and the final audience language before release. Confirm that essential media opens and that the end of the activity provides understandable feedback or clear participation instructions. When material is public, inspect it while logged out. When material is assessed, test the relevant learner access rather than only the editor.
After use, review representative outcomes and student questions. Correct recurring sources of confusion in a controlled later edition and be careful comparing groups exposed to different supports. The completed project should be a reliable learning experience with a visible path for improvement, not merely an impressive count of translated words.
Frequently asked questions about Formative translation
Does one language setting translate everything?
No. Interface controls, teacher-authored material, external links and user contributions may be handled differently. Identify the exact feature’s scope and test each part of the learner journey instead of assuming a single setting covers every source.
Is browser or automated translation safe for assessments?
It depends on the construct. Assistance may improve access when the subject is not language knowledge, but it may reveal the answer during a vocabulary test. Specify allowed tools and review the target-language answer criteria before using scores.
Must the translated lesson use identical words?
No. Different languages often require changed syntax, longer phrases and more explanation. Preserve the intended information, reasoning and learning demand. A natural equivalent is often more faithful than a literal sentence that confuses the audience.
What if a translated option changes the correct answer?
Treat it as a substantive assessment defect. Solve the destination-language item independently, revise the choices or the wording and check feedback and scoring again. Do not rely on the source-language key as authority after translation.
How can learners report a problem?
Provide a straightforward feedback route. Ask what they saw, what they thought the instruction meant and what action followed. Compare the original with the actual displayed version and fix the source of confusion without dismissing the learner’s evidence.
Do captions count as translation?
Not always. Captions may transcribe speech in the original language; translation changes the language. Both can contain errors in names, technical terms and negation. Verify the relevant source audio or transcript before treating caption wording as evidence.
How should we compare multilingual results?
Check the skill measured, source material, difficulty, available assistance and scoring. Where conditions differ, explain that difference. A score alone does not establish equivalent knowledge or a fair comparison.
Do students need the source-language version?
Sometimes it is helpful for verification and attribution; sometimes assessment rules restrict access to it. Decide according to the purpose and preserve the authoritative original for teacher review, even when learners primarily use the translated edition.
How do privacy and permissions affect localization?
A new audience may have a different access role and new safeguarding obligations. Review public links, participant names and the ability to edit or comment. Translation should not become an accidental invitation to publish private student work.
Where should a teaching team start?
Choose a single activity with a clear learning objective. Make the entire path usable, test it independently and document meaningful changes. A small dependable example establishes an approach for larger multilingual collections.
Official guidance and documentation
Product features, interface labels and subscriptions can change. Verify the applicable feature in the current institution account before relying on it for a class or publishing a claim about its availability. These official documentation pages support the particular product distinctions in this guide.
- Formative language settings
- Building an activity
- Multi-part questions
- Teacher-paced mode
- Multiple-choice questions
Related guides to multilingual educational tools
Other platforms face similar questions about language, yet their answer rules, source materials and permissions may be different. The following guides offer distinct product-specific examples without treating every classroom tool as the same translation project.
What a good multilingual learning experience achieves
The final measure is whether a learner can recognize the purpose of an activity, access the information needed, perform the intended intellectual work and understand what the result says. In Formative, as in any educational platform, a finished translation is not simply another language displayed on the screen. It is an experience whose meaning, assistance, sources and outcomes have been checked together.
Begin with one useful lesson, make thoughtful changes and verify them from the learner’s point of view. Keep what the original teaches, improve what the new audience needs, and describe limits honestly. That is how translation becomes a practical part of good teaching rather than a decorative last step.