Have AI plant why-questions inside the worked example

Students who explain a worked example to themselves learn far more from it than students who just read it, and prompting works almost as well as doing it spontaneously: Chi et al. (1994) found that students prompted to self-explain significantly outperformed controls on both immediate and transfer tests. The craft is in the placement. Questions belong at the two or three steps where the reasoning is least obvious, and they must ask why, not what. (Manning's self-explanation-prompt-designer skill turns this placement work into a two-minute prompt.)

In practice

Here is a worked example for Year 10 maths:
[paste worked example]
Target understanding: [the insight, not the steps]

Insert self-explanation questions at the 2-3 steps
where the reasoning is least obvious. For each one:
1. A question asking WHY the step works, not what
   happens in it.
2. The deep answer a student should reach.
3. The shallow answer that signals mere describing,
   with one follow-up question to push past it.
Do not add questions at routine steps.

• Paste the full worked example and name the target understanding, then ask for questions at only the two or three conceptually dense steps, never at routine ones.
• Demand why-questions that target reasoning (why add 9, not 16?) and reject what-questions that only elicit a description of the step.
• Ask for the shallow answer to expect at each question, plus a follow-up that pushes a describing student back to the reasoning.
• Print the questions in the margin or read them aloud at the board; the format matters far less than the placement.