Methods · Creation, transfer & long horizon

Research question refinement

Turn a vague interest into a question that can actually be answered in the space and time you have.

When to use this: Before you start studying
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

Your topic is still a vibe (“something about cities and health”) and a deadline is real.
A supervisor keeps asking what you would count as an answer.
You can name an area but not a question that fits your time and methods.

What to ask AI

Type something like: “I'm interested in [vague area]. Don't suggest a topic. Ask me questions until we find what I'm actually curious about, then help me state one answerable question.”

Ask what would count as evidence, and what is out of scope.

Keep the hard part on your side: you answer the clarifying questions; the AI does not hand you a fashionable topic.

How it helps

Research question refinement turns a vague interest into a question you can actually answer with the time, data, and methods you have.

Vague interests feel productive because they never fail. Answerable questions can fail, that is the point.

The AI’s job is to interview you until the question has a clear object, a feasible method, and a sense of what would count as an answer.

If you leave with a polished question you cannot operationalise, you only polished the wording.

How to run it

State the vague interest in one sentence without trying to sound academic.
Let the AI ask one clarifying question at a time about what you care about and what you can observe.
When a candidate question appears, say what evidence would answer it.
Cut scope until the question fits your timeline and skills.
Write the final question and the out-of-scope list side by side.
Stop when you can explain both to a friend without hand-waving.

Example exchange

You

I'm interested in social media and teenage sleep. Don't suggest a topic. Interview me until the question is sharp enough for a term project.

When you say social media, what behaviour do you actually mean: time on apps, night-time use, or content type?

You

Night-time use after 10pm, mostly.

Good. What would count as an answer you can actually deliver: a correlation from a survey you can run, or a causal claim you cannot support yet?

Copyable prompt

I'm interested in [vague area]. Don't suggest a topic. Ask me questions until we find
what I'm actually curious about, then help me state it as a single answerable question.
Also ask what would count as evidence and what is out of scope.

The Tell

If you can recite a polished question but cannot say what would count as an answer, the wording got refined and the inquiry did not.

If the AI gifted you a trendy topic you feel no curiosity about, you practised compliance, not refinement.

Principle evidence

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

The underlying learning idea is rated moderate. Sharpening inquiry questions is standard research-methods and classroom inquiry practice. Clear, answerable questions improve study design; the tradition is pedagogical more than a single landmark experiment.

AI delivery evidence

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

AI as Socratic interviewer for question framing is plausible and lightly evidenced in tutoring research generally; specific gains for AI-led research-question workshops remain speculative. Judge success by whether you can operationalise the question without the chat.