Writing a Research Question with AI

A research question directs what evidence to seek and what method could answer it. Broad prompts such as 'Is AI good for education?' hide multiple populations, outcomes and value judgments. AI can propose narrower versions and expose ambiguous terms, but the researcher must decide which problem matters and what can be studied ethically within the available time. Begin with the assignment, field and audience, then describe the phenomenon in plain language.

For some clinical effectiveness questions, the National Library of Medicine teaches PICO: patient or problem, intervention, comparison and outcome. That structure helps specify a searchable comparison, but it is not the required form for every historical, qualitative or exploratory inquiry. A qualitative question may ask how participants experience a process, while a history question may focus on interpretation of primary sources. Ask AI to identify which parts of a draft question are vague, then choose a structure appropriate to the method.

Test feasibility. Search the key terms in relevant scholarly databases, see what kinds of evidence exist and note where terms have multiple meanings. A question can be precise but still unanswerable if records are inaccessible, the proposed measure is undefined or the requested causal inference is unsupported by the design. Refine the scope without changing the core interest. Ask AI for counterexamples: what evidence would fail to answer the question, and what assumptions are hidden?

Write a final question that can guide selection of sources and a method. Check it with an instructor or supervisor, particularly when people, sensitive data or institutional approval are involved. Keep a record of why wording changed and what evidence the project can realistically collect. AI is useful for iteration and search terms, while the researcher owns the judgment about importance, ethics and answerability.