Worked-example annotation

Label every decision point in a solution with why, so worked examples become training in choice, not in copying steps.

When to use this: First time through the material
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

You ‘understand’ a worked example until the numbers change.
Your notes restate what each step is, not why that move beat an alternative.
You need to turn examples into decision training before near-twin practice.

What to ask AI

Type something like: “Here’s a complete solution I’ve annotated with why at each step plus one rejected alternative. Critique the annotations only, then blank the whys and quiz me.”

Forbid the AI from rewriting the solution.

After annotations, cover the whys and reconstruct them; then solve a near-twin.

How it helps

Unannotated examples teach recognition of a finished path.

Annotation asks: why this move, why not the alternative?

That is what transfers to new problems.

How to run it

Get a complete correct solution.
At each step, write why plus one rejected alternative.
Let the AI critique missing decisions or wrong rationales.
Cover the annotations and reconstruct the whys.
Solve a near-twin problem cold.

Example exchange

You

I’ve pasted a worked related-rates cone solution and annotated each step with why I did it, plus one alternative I rejected. Can you check the annotations, not the algebra?

Send the annotations for the steps you’re least sure about.

You

Step 3: I differentiated both sides with respect to time because radius and height both change. I rejected differentiating only h. Step 4: I substituted the given rates…

Step 4 is thin. Why substitute before isolating the unknown rate?

You

Honestly I’m not sure. I just followed the solution order. Similar triangles came in earlier to relate r and h; I almost used the wrong proportion.

Mark that similar-triangles choice clearly. Cover your whys: why similar triangles, and why not treat r as constant?

Copyable prompt

I’ve pasted this solution: [ ].
I’ll annotate each step with why I did it and one alternative I rejected.
Critique my annotations only, then blank the whys and quiz me.

The Tell

You know this method has flipped when you annotated once and cannot justify steps on a fresh problem.

Restating what a step is (not why) is copying in disguise.

The near-twin without notes is the real test.

Principle evidence

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

The underlying learning idea is rated strong. Labelling decision points in worked examples is self-explanation on examples (Chi; Renkl) paired with the worked-example effect for novices in cognitive-load and example-based learning research.

AI delivery evidence

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

AI annotation quality is speculative. The method works when you write the whys; chatbot-generated annotations you only read are weak delivery of the same principle.