A21

Backwards scaffolding

Start from a finished answer and reconstruct the reasoning backwards, so you see why each prior step was forced, not just how to mimic a forward template.

When to use this: While you practise
Learning impact: Reliable gain
Learning evidence: Credible with caveats
Where you are with the topic: Building understanding
How often to use it: Worth repeating often

When to use it

You have twenty minutes tonight and need real practice, not more highlighting.
Homework is done, but you still want one honest attempt before bed.

What to ask AI

Type something like: “Here is a complete correct solution. Do not explain it forwards. Hide all but”.

Tell it to ask one question at a time and wait for your answer before continuing.

Keep the hard part on your side: you retrieve, attempt, explain, or decide before you ask for help.

How it helps

Worked examples taught only forwards train pattern matching: "when I see X, do Y." That works until the problem is dressed differently.

Backwards scaffolding asks a different question: given that this is the answer, what had to be true just before, and before that? You rebuild the dependency chain from the end.

This is powerful for proofs, derivations, design rationales, and essay arguments, anywhere the destination is known and the path is the learning.

It complements Worked-example study (forward worked-example study) and Completion problems (completion).

Use forward to see a legal path; use backward to see necessity.

How to run it

Take a finished correct solution (yours after marking, or a trusted model).
Next: cover all but the last line. Ask: what must the previous line have been, and why?
Next: step upward one reveal at a time, you propose, then check.
Next: at each step name the justification ("by chain rule," "by intermediate value theorem," "because the burden was on the claimant").
Next: when stuck, ask for the category of move, not the line (Contingent hinting habits).
Next: finish by running forward once without notes. Backward builds structure; forward confirms you can perform.

Example exchange

You

Here's a complete correct proof. Don't explain it forwards. Hide all but the last two lines, and ask me what must have been true just before them.

Last two lines shown. What claim had to be established immediately before this conclusion?

You

That n is divisible by both 2 and 3, so divisible by 6?

Almost, check whether your course treats that step as already proved or as needing Euclid's lemma. Say which, then I'll reveal one more line upward.

Copyable prompt

Here is a complete correct solution. Do not explain it forwards. Hide all but
the final result and ask me what the previous step must have been and why. Wait
for my answer before revealing. Continue stepping backward to the start.
SOLUTION: [ ]

The Tell

Here is how you know this method has flipped: you can recite the solution forwards from memory and cannot say, for a mid-line, what would go wrong if it were omitted.

In that moment the AI (or the schedule, or the story) did the thinking, and you only recognised a finished product.

Tighten the prompt for backwards scaffolding so you attempt, decide, or retrieve before anything is handed to you.

Principle evidence

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

The underlying learning idea is rated moderate. Working backwards from a finished answer relates to means-ends analysis caveats in CLT (novices overloaded by search) and to worked-example study of solution structure, here you reconstructs the path. Also echoes proof/strategy teaching (backward chaining). Principle moderate; AI-guided backward reconstruction speculative.

AI delivery evidence

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

The learning principle may be moderate, but that does not prove an AI session delivers it well. Treat AI delivery as speculative unless a study of tutoring with this method is named, and none is claimed here.