Methods · Generative practice & retrieval

Error-specific remediation

Get a micro-lesson aimed only at the misconception your answer revealed, not a re-teach of the whole chapter.

When to use this: While you practise
Learning impact: Strong change
Learning evidence: Strong research tradition
Where you are with the topic: Building understanding
How often to use it: Worth repeating often

When to use it

Feedback named your error, and you need a short repair focused on that belief.
Re-reading the chapter feels huge compared with the precise hole you just hit.
You keep missing items for the same underlying reason.
You want a tiny lesson, one check question, then a return to practice.

What to ask AI

Type something like: “My error pattern is [description]. Teach only that misconception in a short micro-lesson. Then one check question. No full chapter review.”

Paste the wrong working so the AI targets your belief, not a generic topic summary.

Ask it to stop after the check question until you answer.

How it helps

A misconception is a specific wrong belief that produces a family of errors.

Re-teaching everything dilutes attention and often misses the belief that actually failed.

Error-specific remediation is a scalpel: short explanation, one check, back to practice.

It follows targeted feedback; it does not replace the attempt that revealed the error.

How to run it

Capture the error in one sentence (“I drop the inner derivative in chain rule”).
Ask for a micro-lesson aimed only at that belief.
Answer one check question without notes.
If you pass, request a fresh practice item on the same skill.
If you fail the check, ask for a second angle on the same misconception, still not a chapter tour.

Example exchange

You

My recurring mistake is applying the power rule to a composition and forgetting to multiply by the inner derivative. Can you give a short micro-lesson on that error only, then one check question?

When a function is f(g(x)), the rate of change is how fast f changes at g(x), times how fast g changes. The second factor is easy to drop because the outer form looks finished. Check: differentiate (4x+1)³ and say aloud what the inner derivative is.

You

3(4x+1)² times 4, the inner derivative is 4.

Clear. Next I will give one new problem only, same trap pattern, still no full solution until you attempt.

Copyable prompt

My error pattern: [one-sentence misconception].
Give a short micro-lesson aimed ONLY at that belief, then ONE check question.
Do not re-teach the whole topic. Wait for my check answer before any more practice.

The Tell

You will know the method has flipped when remediation becomes a fresh lecture on the entire topic, or you skip the check question and call yourself fixed.

Then you received content, not a repair.

Keep the lesson tiny, answer the check cold, and only then return to practice problems.

Principle evidence

Strength of the underlying learning idea, not a claim about AI products.

Formative assessment aims instruction at diagnosed needs. Remediation tied to a specific misconception is more efficient than wholesale re-teaching when the failure is local.

AI delivery evidence

Whether an AI tutor delivers this method well is a separate question.

AI may mis-identify your misconception. If the micro-lesson feels off-target, correct the diagnosis in plain language before accepting more explanation.