# Fade help when performance earns it

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

---

## Fade help when performance earns it

ID: MHC-D-RESEARCH-9004 · Version: 0.1.0 · Kind: protocol
Source: https://vedokrok.com/knowledge/fade-help-when-performance-earns-it

Support should leave because evidence changed, not because the lesson reached page ten.

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

### Use when

- When a learner succeeds with examples, hints or AI assistance but independent performance is the real goal.

### Avoid when

- Prior knowledge is domain-specific. Independence in one class of cases does not justify removing support from a new or higher-risk task.

### Explanation

Begin with enough guidance to produce a correct model of the task. Remove one layer of assistance after repeated accurate use of the relevant decision or procedure, then check independent performance on a fresh case. If the check collapses, restore the specific missing support rather than restarting everything.

### Steps

1. Identify the assistance currently carrying part of the task.
2. Check which step the learner can already perform correctly.
3. Remove one support layer on a fresh case.
4. Restore only the missing support if performance breaks.

### Example

An AI coding assistant first suggests the test structure; later the developer writes the test plan unaided and uses the assistant only to critique edge cases.

### Check

The learner can perform the targeted step on a fresh case without the removed support.

### Limits

- Prior knowledge is domain-specific. Independence in one class of cases does not justify removing support from a new or higher-risk task.

### Evidence and sources

- supports: A 2025 meta-analysis found that lower-prior-knowledge learners benefited more from higher instructional assistance, while higher-prior-knowledge learners benefited more from lower assistance. — RS-6958B9284DD609A1. The effect is moderated by domain and educational status, and does not prescribe a single assistance threshold. (Abstract and highlights)
- supports: A meta-analysis of adaptive training found that effects varied by instructional intervention, with adaptive difficulty showing the strongest results among the examined categories. — RS-05CB28C0C7CF56B3. The literature was heterogeneous and much of the context was military training; this does not validate every adaptive-learning product. (Abstract)
- RS-6958B9284DD609A1: A cornerstone of adaptivity – A meta-analysis of the expertise reversal effect — https://www.sciencedirect.com/science/article/pii/S0959475225000660
- RS-05CB28C0C7CF56B3: Adaptive training instructional interventions: A meta-analysis — https://pubmed.ncbi.nlm.nih.gov/39083372/

No review details supplied.
Related (compare_with): https://vedokrok.com/knowledge/restore-help-when-errors-stop-teaching-you-anything
