research-brainstorming

Brainstorming Research Questions Into Rigorous Designs

Help turn research ideas into fully formed, pre-registered study designs through natural collaborative dialogue.

Start by understanding the current project context, then ask questions one at a time to refine the research question. Once you understand what you're studying, present the design and get user approval.

Anti-Pattern: "This Is Too Simple To Need A Design"

Every study goes through this process. A correlation check, a single model comparison, a parameter sweep — all of them. "Simple" analyses are where unstated hypotheses and post-hoc rationalization cause the most damage. The design can be short (a few sentences for truly simple studies), but you MUST present it and get approval.

Checklist

You MUST create a task for each of these items and complete them in order:

1. Explore project context — check existing papers, data, prior results, research notes. For data specifically: identify the exact source, version, and preprocessing state available. If the user is unsure which version of the data exists, surface this gap before proceeding.
2. Ask clarifying questions — one at a time, understand the research question, constraints, and what success looks like
3. Examine the literature gap (with Devil's Advocate) — structured search for what has NOT been done, plus active search for contradictory evidence
4. Elicit the null hypothesis — exact statement of H0, not just "no effect"
5. Test falsifiability — what specific result would DISPROVE the hypothesis?
6. Enumerate confounds (including data leakage) — list the 3 most likely confounding variables AND check for data leakage risks (temporal, subject-level, preprocessing, feature). See docs/references/data-checklist.md §3 for the leakage taxonomy.
7. Assess statistical power — expected effect size, required sample size, power justification
8. Pre-specify the primary outcome — exactly one primary outcome measure, chosen before seeing results
9. Propose 2-3 study designs — with trade-offs and your recommendation
10. Present the research design — in sections, get approval after each section
11. Write the research design document — save to docs/eureka/designs/YYYY-MM-DD-<topic>-design.md
12. Design self-review — check for placeholders, internal consistency, scope, ambiguity, and verify that null hypothesis and falsifiability criterion both exist
13. Dispatch design-document-reviewer subagent — fresh-eyes review of the design document against the 11 mandatory questions (9 scientific + 2 narrative framing) and placeholder/consistency checks. Block progression on Issues Found; fix and re-dispatch until Approved
14. User reviews written design — ask user to review before proceeding
15. Transition — invoke eureka:hypothesis-first to register the hypothesis