---
name: research-question
description: Turn a research interest or unresolved finding into a bounded, answerable question with explicit concepts, scope, alternatives, and feasibility criteria.
---

# Research question

Use this Academic Research Assistant skill when a project needs a defensible question before selecting a design or collecting evidence. Follow the [research contract](../../docs/RESEARCH-CONTRACT.md). A question may be descriptive, explanatory, predictive, interpretive, evaluative, or mathematical; do not force every project into a treatment-versus-control format.

## Inputs and initial decisions

Collect the topic, intended contribution, audience, disciplinary context, available evidence or materials, time and resource limits, and any fixed constraints. Ask for missing information only when it changes the question materially. With a broad starting point, produce a small set of meaningfully different candidates and state the assumptions used. Separate the user's motivation from an empirical premise that still needs verification.

Identify the unit of analysis and the level at which the claim will be made. A study of individuals cannot automatically establish a property of institutions; a simulation of a model cannot by itself establish the behavior of a physical population. Distinguish the object being studied, the quantity or interpretation sought, and the conditions under which an answer would apply.

## Develop and challenge the question

1. Rewrite the interest as an uncertainty: what is unknown, disputed, poorly bounded, or insufficiently explained? Record the evidence currently supporting that characterization. If the literature has not been searched, call the gap provisional and route verification to [literature search](../literature-search/SKILL.md).
2. Define the central concepts and competing meanings. For a measurable construct, propose observable indicators and explain what they omit. For an interpretive concept, describe whose meanings and which contexts matter. For a formal problem, specify mathematical objects, assumptions, and the desired statement.
3. Bound the scope using relevant dimensions: population, setting, period, parameter range, corpus, scale, or computational regime. Explain why each bound matters, rather than adding arbitrary restrictions for convenience.
4. State plausible alternative answers. If every conceivable result would support the proposed interpretation, revise the question or explain why an exploratory approach is appropriate. A useful exploratory question can remain open while specifying how observations will constrain later claims.
5. Map each candidate to evidence that could answer it. Identify access, measurement quality, ethical constraints, required expertise, and a minimum viable study. Reject or narrow candidates whose decisive evidence cannot realistically be obtained.
6. Test alignment: would the proposed evidence actually answer the wording? Association questions can use observational comparisons; causal questions need a credible identification strategy; prediction requires evaluation on cases not used to fit the model.

## Adapt to the research mode

For quantitative work, define the target quantity or estimand, comparison, outcome, timing, and relevant uncertainty. Distinguish a substantive research question from a statistical null hypothesis. Avoid defining success solely as a small p-value.

For qualitative work, name the experience, process, practice, or meaning being explored, its context, and the standpoint from which it will be studied. Do not embed a preferred explanation in interview wording. Allow a documented revision of the question as engagement with the field deepens.

For theoretical work, separate a conjecture, derivation, classification, existence question, and explanatory framework. State what would count as a counterexample or a failure of the proposed argument. For computational work, specify whether the goal concerns an algorithm, a numerical approximation, a software artifact, or a modeled system; these require different evidence.

## Deliverable and verification

Return a question brief containing: the selected question; plain-language meaning; scope and definitions; motivation with verified source locations; alternatives; proposed evidence; feasibility risks; and explicit exclusions. If several candidates remain, compare them on answerability, significance, evidence access, and expected uncertainty without pretending a numerical score makes the decision objective.

Verify every factual premise against available sources and label unsupported premises as assumptions. Do not guarantee originality or publishability. End with concrete conditions under which the question should be narrowed, revised, or abandoned. Hand the accepted brief to [methodology design](../methodology-design/SKILL.md), preserving unresolved decisions instead of silently choosing them.