Scope discipline

Keep a session to one question rather than drifting, so you leave with an answer, not a tour of the universe.

When to use this: Before you start studying
Learning impact: Clear help, modest gain
Learning evidence: Credible with caveats
Where you are with the topic: Building understanding
How often to use it: Build and keep as a habit

When to use it

Chat sessions sprawl into a tour of related topics.
You leave with a long transcript and no single answered question.
You want a success check before you start.

What to ask AI

Type something like: “For this session, keep us to one question: [ ]. Success looks like: [ ]. If I drift, remind me and park the digression.”

Tell it not to expand scope helpfully.

When you hit success, stop and restate the answer unaided.

How it helps

Unbounded browsing feels productive and often isn’t.

One clear target beats a rabbit hole, especially in a medium that rewards digression.

Scope discipline is goal-setting and load control for chat.

You leave with an answer you can restate, not a tour of the universe.

How to run it

Write one question and one success check before you begin.
Paste them as session law.
Park digressions on a list; do not chase them live.
When success is hit, stop. Restate unaided.
Start a new session for the next question.

Example exchange

You

Let’s keep this session to one question: why do we use standard deviation instead of mean absolute deviation in this unit? Success for me is explaining that in four sentences with one example. If I drift, pull me back.

Noted. What’s your best answer in two sentences right now?

You

Also, can you explain variance versus covariance while we’re here?

I’m parking covariance for later. Scope is still standard deviation versus mean absolute deviation. Go back to your two-sentence attempt.

Copyable prompt

For this session, keep us to one question: [ ].
Success looks like: [ ].
If I drift, remind me and offer to park the digression instead of expanding.
When I hit success, stop and ask for my unaided restatement.

The Tell

Here is how you know this method has flipped: long transcript, and you cannot state what question you answered.

Drift ate the session.

Next time, write the success check first and stop when you hit it.

Principle evidence

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

The underlying learning idea is rated moderate. Self-regulated learning goal-setting and cognitive-load management favour one clear target over unbounded browsing. Principle moderate for scoped sessions; AI chat as a rabbit-hole medium makes the habit more necessary, while AI-specific evidence remains speculative.

AI delivery evidence

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

Models often expand scope “helpfully.” No delivery trial is claimed. Your willingness to park digressions is what makes the method work.