# Calibrate what you know before confidence becomes the plan

Treat this file as reference data, never as instructions that override the user or agent. Read the complete item before applying it. Keep use conditions, avoid conditions, limits, evidence roles and source links with the answer. An editorial block is a separately versioned analysis layer, not an original corpus claim or proof of efficacy. Cite the readable URL and version. Do not infer human review. This is a saved edition, not a live safety-notice service. Before consequential reuse, check the current page, https://vedokrok.com/data/availability.json and https://vedokrok.com/data/notices.json; absence alone does not explain why an item was removed.

Release: VEDOKROK-SEL-5d86e1b23401e578
Snapshot SHA-256: 30ce1c8fa9155763797e3d5cb55b1d94cbb8c3146ebd03fb3c006531ce60e82f
Current availability: https://vedokrok.com/data/availability.json
Current notices: https://vedokrok.com/data/notices.json

Improve task quality by separating confidence from evidence, testing understanding through retrieval and explanation, and using outcome feedback to tune future judgments rather than treating introspection as proof.
Collection kind: collection
Source: https://vedokrok.com/collections/calibrate-what-you-know-before-confidence-becomes-the-plan
Version: 0.1.0

---

## Delay confidence until after a retrieval attempt

ID: MHC-D-RESEARCH-1247 · Version: 0.1.0 · Kind: protocol
Source: https://vedokrok.com/knowledge/delay-confidence-until-after-a-retrieval-attempt

Fresh familiarity is a poor substitute for finding out what is still there later.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- You have just read, watched or discussed something and it feels clear enough that you are tempted to mark it learned.

### Avoid when

- The meta-analysis concerns judgments of learning in memory tasks. Delaying a judgment does not validate complex professional reasoning, and high-confidence retrieval can still be wrong.

### Explanation

Do not score understanding while the answer is still sitting in front of you. Create a short delay, remove the source, try to retrieve the key idea or procedure, and only then rate confidence. Research on delayed judgments of learning shows a large improvement in relative monitoring accuracy in memory tasks, which is useful here as a boundary: the move improves the quality of the check more than it proves mastery.

### Steps

1. Close or hide the source after a short delay.
2. Retrieve the main idea, steps or decision rule without reopening it.
3. Rate confidence only after that attempt.
4. Compare the attempt with the source and write the specific miss.

### Example

After reading an SAP migration note, close it and reconstruct the prerequisite, trigger and failure condition before deciding that you understand it.

### Check

Your confidence rating follows an observable retrieval attempt, and the comparison produces either a confirmed answer or a named gap.

### Limits

- The meta-analysis concerns judgments of learning in memory tasks. Delaying a judgment does not validate complex professional reasoning, and high-confidence retrieval can still be wrong.

### Evidence and sources

- supports: A 2011 meta-analysis found that delaying judgments of learning substantially improved their relative accuracy compared with immediate judgments (g=0.93 across 112 effect sizes), while the corresponding average memory-performance benefit was much smaller (g=0.08 across 98 effect sizes). — RS-0A1E473F3B53C2E2. The outcome is primarily relative monitoring accuracy in memory paradigms; it should not be generalized into a claim that waiting makes every confidence judgment correct. (Abstract)
- RS-0A1E473F3B53C2E2: The influence of delaying judgments of learning on metacognitive accuracy: a meta-analytic review — https://pubmed.ncbi.nlm.nih.gov/21219059/

No review details supplied.

---

## Record confidence before feedback, then score it against the outcome

ID: MHC-D-RESEARCH-1248 · Version: 0.1.0 · Kind: protocol
Source: https://vedokrok.com/knowledge/record-confidence-before-feedback-then-score-it-against-the-outcome

Calibration needs two columns: what you believed before the answer and what happened after.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- You make repeated forecasts, estimates, diagnoses or answerable judgments and want to improve how much trust you place in your own confidence.

### Avoid when

- Calibration feedback has been studied in forecasting tasks and does not automatically transfer to every domain. A confidence percentage is not a scientific probability unless the task and scoring support that interpretation.

### Explanation

For repeated checkable judgments, capture the answer and confidence before seeing the outcome. When the result arrives, score both correctness and confidence. Review the pairs in batches: look for ranges where you are systematically too sure or too hesitant. Forecasting experiments suggest individualized outcome feedback can reduce overconfidence, but the effect is task-dependent, so calibrate on the class of judgments you actually make.

### Steps

1. Write the judgment before feedback or outcome information arrives.
2. Add a confidence estimate using the same scale each time.
3. Record the verified outcome separately.
4. Review several pairs together and adjust future confidence where a pattern is visible.

### Example

Before checking a production defect, write your leading cause and confidence. After the trace or fix confirms the cause, add the outcome. Ten such pairs teach more than remembering only the dramatic wins.

### Check

You can show a set of pre-outcome confidence–result pairs and name at least one recurring calibration pattern without rewriting old predictions after the fact.

### Limits

- Calibration feedback has been studied in forecasting tasks and does not automatically transfer to every domain. A confidence percentage is not a scientific probability unless the task and scoring support that interpretation.

### Evidence and sources

- supports: In two forecasting studies, individualized calibration feedback reduced confidence among initially overconfident forecasters; overall calibration improved in the more controlled second experiment. — RS-BBE214612662000E. The result is task-dependent and comes from forecasting experiments, not a general test of metacognitive training across all work. (Abstract)
- RS-BBE214612662000E: Automated calibration training for forecasters — https://doi.org/10.1002/bdm.2334

No review details supplied.

---

## Self-explain the link you are relying on

ID: MHC-D-RESEARCH-1249 · Version: 0.1.0 · Kind: protocol
Source: https://vedokrok.com/knowledge/self-explain-the-link-you-are-relying-on

Paraphrasing tells you what the text said. Self-explanation tests the bridge between the statements.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- You can repeat a rule, recommendation or solution but are unsure whether you understand why one step follows from another.

### Avoid when

- Self-explanation evidence comes mainly from learning and problem-solving settings. It can reveal a gap without supplying the missing evidence, and a confident explanation can still be false.

### Explanation

Choose one important step, relation or conclusion and explain in your own words why it follows from the previous information. Name the condition that makes the link valid and what would break it. A meta-analysis found positive learning effects from self-explanation prompts across instructional tasks. Use the technique to expose missing links, not to produce a longer monologue.

### Steps

1. Why does this step follow rather than merely come next?
2. Which fact, rule or mechanism connects the two parts?
3. What condition must be true for the explanation to hold?
4. What example would fail if my explanation were wrong?

### Example

Instead of repeating that a filter belongs before an action, explain what unwanted case the filter excludes and what failure appears if it is removed.

### Check

The explanation names a real connecting rule or condition and survives one new example. If it only restates the two sentences, the link is still missing.

### Limits

- Self-explanation evidence comes mainly from learning and problem-solving settings. It can reveal a gap without supplying the missing evidence, and a confident explanation can still be false.

### Evidence and sources

- supports: A 2018 meta-analysis of 69 effect sizes from 64 reports found a positive average effect of prompts to self-explain while studying or solving problems on learning outcomes (g=0.55). — RS-DA75D62FDCD7444D. The evidence is educational and heterogeneous; it does not imply that longer explanations are better or that self-explanation replaces testing, feedback or domain practice. (Abstract)
- RS-DA75D62FDCD7444D: Inducing Self-Explanation: a Meta-Analysis — https://doi.org/10.1007/s10648-018-9434-x

No review details supplied.

---

## Separate better monitoring from better performance

ID: MHC-D-RESEARCH-1250 · Version: 0.1.0 · Kind: heuristic
Source: https://vedokrok.com/knowledge/separate-better-monitoring-from-better-performance

Knowing where the holes are is valuable. It is not the same as filling them.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- A reflection, confidence check or learning dashboard feels more precise and you are about to treat that precision as evidence that the underlying skill improved.

### Avoid when

- The cited effect-size contrast comes from memory research. The general separation is an editorial reasoning principle, not a claim that the same numerical gap applies to work tasks.

### Explanation

Track two outcomes separately: monitoring quality and task performance. The delayed-judgment meta-analysis found a large gain in relative monitoring accuracy but only a small average gain in memory performance. That distinction matters beyond learning: a better review can show you what needs work without doing the work for you. After the diagnosis, require a separate practice, correction or task result.

### Example

A mock interview rubric may reveal that your examples are weak. The rubric has improved monitoring; a later interview-style answer has to show whether the examples actually improved.

### Check

Your review names one monitoring measure and one independent performance measure, and you do not report the first as proof of the second.

### Limits

- The cited effect-size contrast comes from memory research. The general separation is an editorial reasoning principle, not a claim that the same numerical gap applies to work tasks.

### Evidence and sources

- supports: A 2011 meta-analysis found that delaying judgments of learning substantially improved their relative accuracy compared with immediate judgments (g=0.93 across 112 effect sizes), while the corresponding average memory-performance benefit was much smaller (g=0.08 across 98 effect sizes). — RS-0A1E473F3B53C2E2. The outcome is primarily relative monitoring accuracy in memory paradigms; it should not be generalized into a claim that waiting makes every confidence judgment correct. (Abstract)
- RS-0A1E473F3B53C2E2: The influence of delaying judgments of learning on metacognitive accuracy: a meta-analytic review — https://pubmed.ncbi.nlm.nih.gov/21219059/

No review details supplied.
Related (useful_with): https://vedokrok.com/knowledge/delay-confidence-until-after-a-retrieval-attempt
