An adviser shows you six recommendations that worked and two that did not. Six out of eight is 75 per cent. That arithmetic is correct. Whether it describes the adviser’s whole record is a separate question.
The short answer is to ask how the displayed cases were selected, which recommendations are missing, and what outcome could fairly be checked for each one. Keep withdrawn advice, unimplemented suggestions and unresolved outcomes visible without pretending they all mean failure. A more complete record can improve judgment while still leaving uncertainty.
This guide uses original fictional school and community examples. It is about evaluating a claim of dependable advice, not certifying a person’s wisdom with a score. Begin with the Wisdom subject guide for the wider relationship between evidence, values, care and practical judgment. Here we examine one specific weakness: a remembered track record can be much narrower than the decisions it is used to justify.
eduKateSG · Wisdom
Find your next learning step
Identify the claim, counting unit and displayed cases. Choose the question closest to your task, or follow the complete lesson.
Chapter contents
Define the record · Chapters 1–4
Recover the missing cases · Chapters 5–8
Judge the evidence · Chapters 9–13
Practise a fair review · Chapters 14–17
Continue thoughtfully · Chapters 18–20
CHAPTER 1 OF 20 · Define the record
1. Identify the claim hidden inside the success story
A story about one good recommendation may be true and useful. The problem begins when the story is asked to prove a larger claim, such as this person’s advice usually works for people like us.
That larger statement contains several questions. What counts as advice? What counts as working? Which people and circumstances are relevant? Which time period is being described? A collection of attractive examples does not answer those questions automatically.
Imagine a fictional student mentor showing successful project plans. The mentor may have helped those groups thoughtfully. We should not dismiss the help merely because the record is incomplete. But selecting the next group’s plan requires more than knowing that some previous groups succeeded.
Ask what the display establishes directly. It may establish that six named recommendations were associated with outcomes meeting a stated criterion. It does not yet establish how often all recommendations met that criterion, whether they caused the outcomes, or whether the next group faces similar conditions.
A careful reader can appreciate the examples and still ask for the missing denominator. This is a respectful evidence question. It does not require assuming that the adviser intended to mislead anyone. People naturally remember some events more clearly than others, and informal records can lose detail without deliberate deception.
Before counting successes, decide what the count’s unit is. A conversation can contain several suggestions, one revised recommendation or an open question. Those are not necessarily separate attempts at the same task.
For our fictional record, one recommendation means a dated proposal to one group for one specified decision, with an identifiable intended outcome. Several messages clarifying the same proposal remain part of the same case. A materially different proposal is recorded as a revision linked to that case.
This convention is chosen for the lesson. Another investigation may need a different unit, but it should state the rule before calculating. Otherwise, a long conversation can be counted as many successes when it helps, yet only one attempt when it fails.
Suppose the mentor suggests a morning rehearsal, a shared checklist and a shorter introduction for the same presentation. If the question concerns the complete presentation plan, count one case and preserve its components. If the question specifically concerns rehearsal timing, identify which cases actually included that recommendation.
The unit also needs an owner and a date. Advice repeated after the outcome is known does not become a new successful prediction. A record written beforehand helps distinguish what was proposed from the more polished explanation that people may construct afterwards.
CHAPTER 3 OF 20 · Define the record
3. Choose an outcome that belongs to the original task
Working can mean finishing on time, including every participant, improving accuracy or reducing workload. These goals may conflict. A recommendation can meet one criterion while falling short on another.
In our main example, the fictional mentor’s eligible project advice is judged against a previously stated criterion: the group submits its agreed project by the deadline with all required parts present. This is narrow enough to check from the supplied record. It does not measure the project’s creativity, fairness or learning quality.
If a group misses the deadline but learns something valuable, the learning can be acknowledged without rewriting the original deadline outcome. If a group submits on time by placing an unreasonable burden on one member, the deadline result does not settle whether the advice was wise.
Record intended benefits and relevant burdens separately. A success definition should not become so flexible that every possible outcome can be redescribed as a win. Equally, it should not erase legitimate outcomes that were actually part of the original purpose.
For a learner, ask for the success criterion before revealing the result. Then reveal a mixed outcome and invite two sentences: one about whether the stated criterion was met, another about the wider considerations. This protects both factual clarity and humane judgment.
The fictional mentor displays eight completed project cases. Six met the stated deadline-and-completeness criterion, and two did not. The displayed success proportion is therefore 6 ÷ 8 = 0.75, or 75 per cent.
That result describes the eight displayed cases. Its denominator is not all advice ever given, all groups approached or every project that might have benefited. The count’s scope should remain beside the percentage.
| Displayed case group | Count | Included in the displayed rate? |
|---|---|---|
| Met the stated criterion | 6 | Yes |
| Did not meet the criterion | 2 | Yes |
| Total displayed completed cases | 8 | Denominator |
Now ask how those eight reached the display. Were they every eligible completed case? Were they selected because the mentor could contact the groups? Did groups volunteer their stories? Were memorable examples chosen for a short talk? These possibilities imply different limits.
There is no need to reject a short talk for being short. A speaker can present selected examples honestly by labelling them as examples. The problem is treating the display as a complete performance estimate without checking its selection.
The first repair is often wording: in these eight displayed cases, six met the stated criterion. That sentence is less sweeping than a claim about the next person’s likely success, but it is supported by the evidence actually shown.
CHAPTER 5 OF 20 · Recover the missing cases
5. Expand the record without losing its categories
Suppose the mentor’s dated log contains twenty eligible cases in which the recommendation was implemented. Sixteen now have resolved outcomes. Ten met the criterion and six did not. Four are still awaiting the relevant deadline or confirmation.
The eight displayed cases are a subset of the sixteen resolved cases, not an extra set to add. They contain six of the ten successes and two of the six failures. The remaining resolved cases therefore contain four successes and four failures.
| Implemented cases | Met criterion | Did not meet | Pending | Total |
|---|---|---|---|---|
| Displayed subset | 6 | 2 | 0 | 8 |
| Other resolved cases | 4 | 4 | 0 | 8 |
| Unresolved cases | 0 known | 0 known | 4 | 4 |
| Complete implemented record | 10 | 6 | 4 | 20 |
The resolved-case success proportion is 10 ÷ 16 = 62.5 per cent. The earlier 75 per cent was not a numerical error. It described a selected subset with a different balance of outcomes.
Keep the pending cases separate. Calling all twenty resolved would invent information. Dropping the pending category from the report would make it harder to understand the full implemented record.
This is the central lesson: before disagreeing about a percentage, inspect which cases produced it. Correct calculations can support different conclusions when the denominators refer to different populations or stages.
CHAPTER 6 OF 20 · Recover the missing cases
6. Bound the unresolved result instead of guessing it
The twenty implemented cases contain ten known successes, six known failures and four pending outcomes. What can we say about the eventual success proportion if every pending case will later receive one of those two classifications under the same criterion?
If all four pending cases fail, there will be ten successes out of twenty, or 50 per cent. If all four succeed, there will be fourteen out of twenty, or 70 per cent. The eventual complete-implemented-case proportion lies between those endpoints under the stated assumptions.
This is a logical range, not a confidence interval or a forecast. It does not say that the midpoint is most likely, that all values are equally plausible, or that the mentor’s future performance lies in the same range. It simply keeps the unknown outcomes unknown while showing their maximum effect on this finite record.
Compare that with the displayed 75 per cent. In this example, even the most favourable resolution of all pending cases would not make the twenty-case proportion reach 75 per cent. The selected display therefore cannot describe the eventual proportion for those twenty implemented cases.
The conclusion depends on the record being complete, the criterion staying stable and the pending cases being the only unresolved classifications. If another case is discovered or an existing outcome was misclassified, update the account. Do not turn a conditional bound into certainty about facts that were never established.
CHAPTER 7 OF 20 · Recover the missing cases
7. Keep abandoned suggestions visible without inventing their outcomes
The mentor also made four suggestions that were never implemented. Two groups chose another plan, one project was cancelled, and one suggestion was withdrawn before work began. These belong in an issued-advice record, but their status differs from the twenty implemented cases.
Adding them produces twenty-four issued cases: twenty implemented and four unimplemented. It does not produce four additional observed failures. For an unimplemented plan, the outcome that would have followed implementation is not supplied by this record.
Do not quietly drop these cases either. They may matter for understanding usefulness, feasibility, responsiveness or why people chose not to proceed. A plan that is repeatedly impossible to implement deserves investigation even if its hypothetical outcome cannot be scored.
Use a separate implementation proportion if that answers a real question: twenty of twenty-four issued cases were implemented, approximately 83.3 per cent. That is a description of this fictional record, not proof that implementation was desirable in every case.
The withdrawn suggestion also needs its timing and reason. Withdrawing advice before harm occurs can reflect learning. Withdrawing it after seeing an unfavourable result and then erasing the original recommendation distorts the record. Preserve both the original proposal and the revision so that the sequence remains understandable.
The useful response to an abandoned case is often a question, not a score: what prevented implementation, and does that obstacle matter for the next person considering similar advice?
Advice should be allowed to change when facts change. A record that makes every revision look dishonest would punish responsible learning. A record that erases every earlier version would make it impossible to evaluate what was known and recommended at the time.
Suppose the mentor originally recommends a two-week project timetable. After learning that the group has an earlier school deadline, the mentor proposes a shorter plan. Record the initial timetable, the new information, the revised recommendation and the date of the change.
If the group succeeds under the revised plan, it is reasonable to describe that result accurately. It is not reasonable to present the original two-week recommendation as though it had always accommodated the earlier deadline. The improvement belongs to the sequence.
For one case, you might examine whether the adviser noticed missing information, corrected promptly and helped the group recover. Those are process observations. They are not automatically another independent success to add to the project-outcome count.
This connects with the existing guide on caring for people who relied on a decision that must change. That guide owns the responsibilities around a justified revision. This one owns the integrity of the track record used later to claim dependability.
Both questions matter: did the adviser respond thoughtfully, and does the later account tell the sequence honestly?
CHAPTER 9 OF 20 · Judge the evidence
9. Separate a good result from evidence of a good cause
Even a complete record of outcomes after advice does not, by itself, establish that the advice caused those outcomes. Groups bring their own skills, support, resources and decisions. Some would have succeeded under another plan.
Our fictional log tells us that ten implemented cases met the criterion. It does not supply a comparison showing what those same groups would have done without the mentor’s advice. The causal question therefore remains open.
This limit need not make the log useless. It can reveal implementation problems, missed deadlines, recurring constraints and questions for future inquiry. It can support a more honest description of experience. It just cannot answer every question about the advice’s effect.
Ask learners to compare two claims. The first says that ten of sixteen resolved implemented cases met the criterion. The second says that the mentor caused ten groups to succeed. The first follows from the stated classifications; the second requires additional evidence about alternatives and competing explanations.
A fair evaluation also avoids reversing the error. A failed outcome does not automatically prove bad advice if an unforeseeable event intervened. Examine what could reasonably have been known, whether the recommendation fitted the circumstances and whether its risks were handled appropriately.
Outcome records and process records should inform each other. Neither should be used as a shortcut that removes the need to understand the actual decision.
CHAPTER 10 OF 20 · Judge the evidence
10. Compare advice against a relevant alternative
A success proportion has more practical meaning when the alternative is visible. If most groups already meet a simple deadline without help, a high success rate after advice may not show much added value. A different task with severe constraints could produce a lower rate despite useful assistance.
Imagine two fictional contexts. In one, nearly every group has a stable timetable and extensive support. In another, deadlines change and essential materials arrive late. Comparing advisers solely by their success proportions can confuse the task conditions with the quality of the advice.
Do not invent a comparison group to solve this problem. If the record lacks relevant alternatives, state that limit. You can still ask which features of the cases resemble the next decision and which differ.
For a learner exercise, give two advisers the same displayed percentage but different task descriptions. Ask what additional information would be needed before choosing between them. Useful answers include comparable goals, eligibility, case completeness, implementation conditions and relevant costs.
The purpose is not to demand a controlled experiment before accepting any ordinary suggestion. It is to scale the conclusion to the evidence. A low-stakes idea can be tried with modest monitoring. A consequential recommendation needs stronger, relevant support and the appropriate expertise.
Keep the practical question concrete: what does this record help us decide here, and what does it leave unresolved?
CHAPTER 11 OF 20 · Judge the evidence
11. Count burdens as well as successful cases
A binary outcome count treats every success as one and every failure as zero. That may answer a narrow descriptive question. It does not automatically describe the consequences for the people involved.
Consider an invented evaluation where six successful recommendations each produce a benefit valued at one agreed point, while two failed recommendations each create a burden valued at five points. The selected eight-case record still has a 75 per cent success proportion. Under this explicitly chosen scoring rule, its total is 6 × 1 − 2 × 5 = −4 points.
The points are not an objective measure of human welfare. They are a classroom device showing why frequency and consequences are different dimensions. Another set of legitimate values could produce another assessment, which should be stated rather than hidden.
In real decisions, some burdens should not be reduced casually to points. Exclusion, dignity, safety and responsibilities may require direct consideration under the applicable context. A high numerical success rate does not remove those questions.
Ask who received the benefits and who carried the burdens. If one member repeatedly rescued a group’s deadline through excessive work, the project-level count can conceal that experience. Invite relevant information without forcing private disclosure.
Wisdom includes deciding what deserves attention. A record that counts outcomes accurately becomes more useful when it also preserves the material consequences that the count cannot express.
CHAPTER 12 OF 20 · Judge the evidence
12. Check whether several testimonials describe independent experience
Five people can repeat the same success story without providing five independent cases. A group member, teacher and friend may all describe one project. Their accounts may offer useful perspectives, but the project should not multiply merely because it was retold.
Give the underlying case an identifier and connect the accounts to it. Then distinguish the number of witnesses from the number of recommendations or outcomes. This prevents a popular story from silently increasing the apparent size of the record.
Independence also matters when people copy an interpretation. Several accounts may rely on the same mentor’s summary rather than separate observation. Ask which part each person actually witnessed and what source supports the rest.
This does not make shared testimony worthless. Different participants can identify distinct facts, such as whether instructions were clear or whether one person carried most of the work. Those observations help reconstruct the case while remaining part of the same case.
The existing Feedback Source Calibration guide examines domain fit, evidence access and independence when judging a correction. Here the narrower check is whether repeated stories have been counted as new performance evidence.
For a practical exercise, provide six short testimonials arising from three fictional projects. Ask the learner to group them by event, identify corroborated facts and mark statements that merely repeat a common source. The resulting account should become clearer, not automatically more dismissive.
CHAPTER 13 OF 20 · Judge the evidence
13. Ask for a fuller record without turning inquiry into accusation
The way a question is asked can affect whether useful information becomes available. A person who feels accused may defend the display instead of helping reconstruct its scope. Begin with the claim and the evidence needed to understand it.
A calm request might be: these examples help me see your approach; are they all the eligible cases from this period, or a selection? Follow with the specific missing category, such as pending outcomes, recommendations that changed, or plans that were not implemented.
Avoid demanding every private conversation. A proportionate summary can preserve case counts and categories while protecting identities. Where a decision requires a more formal review, use the appropriate authorised process rather than informal pressure for confidential records.
If the complete record is unavailable, the conclusion should reflect that. You can describe the displayed examples and explain that the overall rate remains unknown. You do not need to assume either hidden excellence or hidden failure.
For students, practise rewriting an accusatory question into an evidence question. Compare you only show your wins with how were the examples selected, and what happened to the other eligible cases? The second invites a checkable answer.
Respectful inquiry does not require accepting an unsupported claim. It keeps the conversation focused on what the record can establish, making room for correction without turning uncertainty into a verdict on someone’s character.
CHAPTER 14 OF 20 · Practise a fair review
14. Build a small prospective advice record
The easiest time to preserve a recommendation’s meaning is before the outcome becomes known. A modest record can identify the question, context, proposal, intended outcome, important assumptions and date for review.
Use a fictional case identifier rather than personal details in a classroom exercise. Record whether the recommendation was implemented, changed, withdrawn or not pursued. When the review date arrives, record the observed outcome and the evidence supporting its classification.
Keep a revision connected to the original entry. If the criterion changes, explain why and retain the earlier criterion. An honest new question is useful; silently replacing an inconvenient old one makes the track record difficult to interpret.
| Field | What it preserves |
|---|---|
| Decision and date | The question before the outcome |
| Recommendation and assumptions | What was actually proposed |
| Intended outcome and review date | What can later be checked |
| Implementation status | Whether the advised action occurred |
| Revisions and reasons | How the advice changed |
| Outcome evidence and unresolved issues | What the record supports now |
This routine is an original educational adaptation, not a formal research protocol. The Center for Open Science’s preregistration guidance explains the research distinction between plans made in advance and work shaped after observing results. Its lifecycle open-science framework also connects plans and outcomes. Those research practices motivate a useful comparison, but they do not certify an informal advice log as a scientific study.
CHAPTER 15 OF 20 · Practise a fair review
15. Try a changed record before reading the answer
Here is an independent fictional case. A community organiser displays five completed recommendations: four met their original criterion and one did not. A dated log then reveals twelve implemented recommendations in the relevant period.
Across those twelve, seven met the criterion, three did not and two remain pending. Separately, three issued recommendations were never implemented. One was withdrawn before its review date; two groups chose different plans.
Calculate the displayed success proportion and the resolved-implemented-case success proportion. Then give the lowest and highest eventual proportion among the twelve implemented cases, assuming only the two pending classifications remain unknown.
Explain why the three unimplemented cases do not automatically become failures in that implemented-case calculation. Also explain why erasing them entirely would leave the issued-advice record incomplete.
Now consider a further claim: because seven groups met the criterion, the organiser’s advice caused seven successes. Identify what the supplied record is missing for that causal conclusion. A correct response should distinguish an observed outcome after implementation from evidence about what would have happened under an alternative.
Finally, draft one respectful question you would ask before using the record to choose a plan for a new group. Make the question specific enough that its answer could change your judgment. Broad suspicion is less useful than identifying the relevant missing information.
CHAPTER 16 OF 20 · Practise a fair review
16. Check the changed record and its limits
The displayed proportion is 4 ÷ 5 = 80 per cent. The resolved implemented record contains seven successes and three failures, so its proportion is 7 ÷ 10 = 70 per cent. Both calculations are correct for their stated sets.
Across all twelve implemented cases, the lower eventual proportion is 7 ÷ 12, approximately 58.3 per cent. The upper is 9 ÷ 12 = 75 per cent. Those endpoints assume that each pending case will eventually be classified under the same criterion and that the rest of the record is complete.
The three unimplemented recommendations make fifteen issued cases in total. Their implementation status is known; the outcomes that would have followed their implementation are not. Treating them all as failed implemented attempts would answer a different, poorly defined question.
| Measure | Calculation | What it describes |
|---|---|---|
| Displayed success | 4/5 = 80% | Selected completed examples |
| Resolved implemented success | 7/10 = 70% | Known implemented outcomes |
| Eventual implemented range | 7/12 to 9/12 | Pending outcomes left open |
| Implementation proportion | 12/15 = 80% | Issued cases put into practice |
The two 80 per cent figures happen to match, but their numerators and denominators mean different things. Matching numbers do not establish matching concepts.
For causation, the record lacks the relevant alternative outcomes and adequate control of competing explanations. A useful next question might ask whether the new group’s deadline, resources and constraints resemble those in the recorded cases. That connects the evidence to the actual decision without promising that history will repeat.
CHAPTER 17 OF 20 · Practise a fair review
17. Choose a proportionate response to incomplete evidence
Incomplete evidence does not always require stopping all action. The response depends on consequences, reversibility, available alternatives and the person’s role. A small educational suggestion can sometimes be tested while keeping the uncertainty visible.
For example, a group might trial a short planning checklist for one meeting and review whether it clarifies tasks. Define what will be observed and avoid presenting the trial as proof that the adviser is universally reliable. The test answers a bounded practical question.
A difficult-to-reverse decision with serious consequences needs stronger, relevant evidence and appropriate expertise. A collection of selected success stories should not substitute for the information or professional process the decision actually requires.
Also consider the cost of waiting. Delaying until every historical case is resolved may be impossible or unhelpful. The responsible move can be to use the best available evidence, state the limit, preserve options where feasible and plan an appropriate review.
This is where the Wisdom root’s wider framework matters. Counting honestly supports judgment, but values, responsibilities and affected people still belong in the decision. A more complete denominator cannot determine every legitimate priority.
The final recommendation should therefore explain both its evidence and its scope: given these conditions, this is the action we can justify now, and this is the information that would make us revise it. That is more useful than either automatic trust or automatic rejection.
CHAPTER 18 OF 20 · Continue thoughtfully
18. Practise repair, stabilisation and extension
The repair route separates a displayed subset from the whole eligible record. Give a learner a small set of case cards and ask which cards produced the advertised percentage. Then reveal the missing cards and ask them to restate the claim.
The stabilisation route adds implementation status, pending outcomes and revisions. Ask the learner to calculate only the measures whose definitions are supplied. An answer identifying an undefined denominator can be better than a confident percentage assembled from incompatible categories.
The extension route examines comparison groups, consequences, dependence between testimonials and changes in the decision context. Introduce one complication at a time, keeping the earlier record available so that the learner can explain exactly why the conclusion changes.
Teachers can assess whether the explanation preserves the original criterion, uses the correct set and treats unknowns honestly. They should not describe the exercise as a complete measurement of a child’s wisdom. The observed achievement is careful reasoning in the supplied case.
For independent transfer, use a different domain such as club planning, study methods or a fictional community event. Avoid medical, legal or financial decisions that require specialist advice beyond the exercise. The learner’s task is to identify the record’s limits, not to impersonate a professional adviser.
End with one practical habit: whenever a success story supports a wider claim, ask what belongs in the full record before deciding how much weight the story deserves.
Does an incomplete record mean someone is dishonest?
No. Memory, privacy, lost records and a short presentation can all limit what is visible. Evaluate the claim’s support without inventing a motive.
Should every abandoned recommendation count as failure?
No. Distinguish implementation, withdrawal, cancellation and observed outcomes. An unimplemented plan may raise feasibility questions while leaving its hypothetical outcome unknown.
Is a logical range a forecast?
No. The ranges here show what unresolved classifications could do to a defined finite record. They are not confidence intervals or probabilities for the next case.
Can a complete log prove the advice caused success?
Not by itself. Completeness improves description. Causation requires evidence addressing relevant alternatives and other explanations.
Should responsible revisions damage a track record?
They should be recorded with their timing and reasons. Prompt correction can be valuable. Erasing the earlier recommendation prevents a fair account of what happened.
What if private records cannot be shared?
Use an authorised, proportionate summary where appropriate and state what cannot be verified. Privacy does not justify pretending that an unseen complete record has been checked.
Is a higher success proportion always wiser?
No. Task difficulty, selection, costs, values and responsibilities matter. A narrow success count cannot settle the whole practical judgment.
CHAPTER 20 OF 20 · Continue thoughtfully
20. Continue through the relevant evidence routes
Judgment, care and correction
The Wisdom root connects the numbers to proportionate action and affected people. Caring for people who relied on a decision that must change examines responsible revision. Feedback Source Calibration develops the separate questions of expertise, evidence access and independence.
Selection and a traceable record
School Scientific Sampling Bias provides the broader sampling route. Open Science develops transparent research practice. These specialist owners remain distinct from our ordinary advice-record case; their methods should be used within their stated scopes.
The Center for Open Science’s Preregistration and Lifecycle Open Science pages were checked on 22 October 2026. They support the research comparison concerning advance plans and connected outcomes. All advice ledgers, classroom conventions, bounds and practical questions here are original illustrations. They do not establish a validated score for wisdom or the effectiveness of any real adviser.
Continue into Education: learning that transfers into life and How X Works to practise moving from a plausible story to a supported explanation. Keep the person and the claim separate: ask fairly, count carefully, and make only the judgment the evidence can bear.