Text guides

Research and sources

Start with a question, keep a record of evidence and show where sources disagree. Ask for help organizing research without treating a generated answer as proof.

12 practical tasks. Each guide includes a sample request and a way to check the answer.

Narrow a research question to something answerableA topic like "remote work" cannot be answered, but "which scheduling practices did these three published studies link to fewer missed handoffs" can. Start from the decision the answer will feed. Then set limits on population, period, place and source type until the evidence you can reach would settle the question.
Build a source log you can check laterA source log pairs every claim with the place that supports it. Capture the author, date, the specific passage and how you reached the material at the moment you read it. A link alone often fails later, because pages change and long documents bury the relevant line.
Separate original evidence from summaries of itSeveral articles can repeat one statement that began in a single place. To find out, follow each citation backward until you reach the original data, document or statement, then record how direct each step is. Three outlets citing one report count as one source.
Compare sources that appear to disagreeTwo sources can seem to conflict while measuring different things. Line up their definitions, time periods, populations and methods first. Some disagreements disappear once these match, and some stay. When one stays, say so plainly instead of averaging the claims.
Make uncertainty visible in a research summaryA summary reads as certain unless you mark otherwise. For each claim, record what supports it, what cuts against it and what is still unknown, then state the confidence in plain words. A reader can then see which statements to rely on and which to treat as open.
Choose what to read first from a research listA long reading list needs an order. Rank each item by how directly it addresses your specific question, what type of evidence it offers and whether its age matters. Base the ranking on the text you have, such as an abstract or table of contents, and mark everything else as unknown.
Prepare questions about a paper's methodsRead the methods section alongside the paper's research question. Note who or what was studied, how observations were collected and what the authors compared. Separate the limitations they state from questions you want to investigate. Mark information absent from the excerpt before judging what the findings support.
Explain research findings to a non-specialist readerPlain language works when each simplification still matches the original. Name the question, say briefly how it was studied, state the main finding at the strength the authors gave it, and say what remains uncertain. Then give the practical meaning, kept within what was found.
Check the context around a quoted statisticBefore reusing a statistic, find the source passage and note what it counts, which group it describes and the period covered. For a percentage, identify the denominator or starting value. Record any missing context so you can decide whether the number supports the sentence you want to write.
Make interview notes useful without identifying participantsRemove direct identifiers before sharing notes for review. Use participant codes and keep the context needed for the research. Then inspect combinations such as a rare job, small town and exact date: they may identify someone even after a name is removed. Treat the result as a draft requiring review.
End a research note with useful open questionsClose the note with the unknowns that matter to its stated purpose. For each, explain which decision it affects, what evidence could answer it and where to begin looking. Separate questions already answered in the note from those that need more work, and assign priorities for the next reader.
Recheck sources before reusing old researchOld research can carry claims that no longer hold. List each claim with its source and the date you relied on, find the current version from the authority that owns it, and note whether the substance changed. Revise only what the new evidence supports.

Related topics