Teach the AI

Teach a deliberately naive AI student and handle its questions, so gaps show up in the questions you cannot answer.

When to use this: While you practise
Learning impact: Strong change
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

Explain-back feels too soft because nothing pushes back.
You want follow-up questions from a beginner’s point of view.
You think you understand until someone asks “but why?” twice.
You are preparing to tutor a peer and need rehearsal.

What to ask AI

Type something like: “Act as a confused beginner on [topic]. I will teach you. Ask one sincere question at a time when something is unclear. Do not suddenly become an expert or correct me unless I ask.”

Tell it to stay naive, no stealth lectures disguised as questions.

Keep answering its questions yourself; do not flip roles mid-way.

How it helps

Teaching forces you to sequence ideas and notice holes when the learner gets stuck.

A naive questioner makes the holes public: you hear the question you cannot answer.

That unanswered question is the study plan, not a failure of performance.

The AI must stay a student. If it becomes a co-teacher, the method collapses.

How to run it

Set the AI as a beginner with a specific confusion level.
Teach in short steps; invite one question at a time.
Answer without looking things up first.
When you cannot answer, stop and note the gap.
Fix the gap from a source, then resume teaching from that point.

Example exchange

You

Pretend you are a beginner. I am going to teach you why we use the chain rule. Ask one sincere question at a time and stay naive.

Okay, if I already know the derivative of sin(x), why is that not enough for sin(3x)?

You

Because the inside is changing too. It is 3x, not just x.

But how do I know what to multiply by? Do not give me a formula sheet, explain it like I am new.

Copyable prompt

Act as a confused beginner learning [TOPIC]. I will teach you.
Ask ONE sincere question at a time when something is unclear. Do not lecture,
correct, or suddenly become expert unless I explicitly ask. Stay naive.

The Tell

You will know the method has flipped when the “beginner” starts delivering the perfect explanation for you, or you ask it to just tell you the answer to its own question.

Then you stopped teaching and started consuming again.

Reset the naive role, and treat an unanswered student question as your next retrieval target.

Principle evidence

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

Learning by teaching and explaining to others is a robust generative activity. Question-driven tutoring rehearsal surfaces gaps that silent review misses.

AI delivery evidence

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

Role fidelity is the weak point: models drift into expertise. Frequent reminders to stay naive matter more than fancy prompts.