Protocol
Fade support when unaided performance becomes reliable
Support should have an exit condition.
Before you use it
- Use when
- Worked examples, prompts, checklists or AI hints helped at first but may now be doing work the learner can do alone.
- Avoid when
- Do not fade safety checks, legal controls or job aids that matter in risky or memory-heavy work. The goal is independent skill where independence matters.
An idea to explore.
Try this idea.
Inspect the example, try the idea in your situation, then check your next move.
Why it matters
Assistance that helps a novice can become redundant as knowledge grows. The 2025 meta-analysis found that high-prior-knowledge learners did better with lower assistance on average. A recent geometry experiment also favored fading and found prior knowledge to matter. Remove help in steps: hide one prompt or solution step, test unaided performance, and restore help only where errors show it is still needed.
Steps
- Name the support currently carrying part of the task.
- Remove one layer: a hint, solved step, template field or AI suggestion.
- Run a comparable task without that layer.
- Keep it removed if quality holds; restore or redesign it if performance breaks.
- Repeat until the learner is independent at the level the task requires.
An example
A language learner starts with sentence frames for speaking. Once the target structure is reliable, remove the frame for some prompts. If performance collapses, bring back a smaller cue.
Check your result
The learner can complete a representative task without the removed support while meeting the same quality criterion.
Keep this limit in mind
- Do not fade safety checks, legal controls or job aids that matter in risky or memory-heavy work. The goal is independent skill where independence matters.
My learning
Read. Recall. Repeat.
Save this idea and return to it on any device.
Connected ideas
Useful withUse AI to expose expert patterns, then test whether the skill transfers
These ideas link to this one. Check their conditions and limits before use.
Useful with this ideaMatch learning support to what the learner already knows
Evidence and sources (2)
A 2025 meta-analysis of 60 experimental studies found that learners with low prior knowledge learned more under high-assistance instruction, while learners with high prior knowledge learned more under low-assistance instruction.
The pooled effects were highly heterogeneous, and assistance covered different instructional designs rather than one universal method.
A cornerstone of adaptivity – A meta-analysis of the expertise reversal effect · Abstract and main results
A 2025 randomized study of 114 sixth-grade students found that a fading condition produced the largest pre-to-post effect sizes among four geometry-learning conditions, and prior knowledge moderated the effect.
The study was small and specific to sixth-grade geometry delivered through one online tutor; it cannot establish a general optimal fading schedule.
Does working memory moderate the effect of fading on math performance? · Abstract
Edition and review
Version 0.2.0 · Published
No review details are supplied for this edition.