Research
Investigate a question and produce an evidence-backed synthesis with visible uncertainty.
Purpose
Investigate a question and produce an evidence-backed synthesis with visible uncertainty.
Recommended
- Turn a broad topic into a concrete research question whose answer could change a decision.
- State the scope, freshness requirement, and source-quality requirement before collecting evidence.
- Prefer primary or first-party sources for factual claims, and record the source and date for each material finding.
- Distinguish sourced facts, interpretation, and inference; preserve relevant competing explanations or conflicting evidence.
- Explain what the evidence does not establish, including context limits, uncertainty, and unresolved gaps.
- End with a concise synthesis and a deliverable that maps conclusions back to the evidence collected.
Avoid
- Filling evidence gaps silently or presenting an inference as a sourced fact.
- Treating a search result, benchmark, anecdote, or single source as universally decisive without checking scope and method.
- Ignoring source freshness when the question concerns changing products, policies, prices, or capabilities.
- Collecting citations without explaining how they support the conclusion.
Prompt skeleton
Prompt skeleton — deterministic, no model call
Research question [State the precise question and the decision or action it informs.] Scope [Define included systems, versions, populations, dates, and exclusions.] Freshness requirements [State the relevant as-of date, update window, or version boundary.] Known context [Record verified background, prior findings, and assumptions before searching.] Source requirements [Specify preferred source types, authority, independence, and citation needs.] Evidence to collect [List the facts, measurements, examples, or counterexamples needed to answer the question.] Competing explanations [Record plausible alternatives and what evidence would distinguish them.] Uncertainty [State evidence gaps, conflicts, limitations, and confidence boundaries.] Synthesis [Summarize the answer, separating sourced facts from inference.] Deliverable [Define the final artifact, audience, citations, and decision-ready conclusion.]
From templates/research.md. Placeholders describe what to fill in; nothing here invents project facts.
Evidence
official
- officialACM SIGSOFT's empirical standards use method-specific checklists and distinguish essential, desirable, and extraordinary attributes, including validity, reliability, objectivity, and reproducibility concerns.
ACM SIGSOFT Empirical Standards for Software Engineering ↗