Judgment in the Age of AI · Article 25
AI can rapidly suggest research topics, keywords and apparently important papers, but it may omit contrary evidence, favour easily accessible English material or invent citations. It is useful for widening the entrance to a search, not for deciding by itself what a field contains.
Build a funnel rather than asking once
- Concept: Ask for synonyms, older terminology, related concepts and exclusion terms.
- Database: Take the search string into specialist databases, library catalogues or scholarly search tools.
- Publication: Verify title, authors, journal, year and DOI.
- Evidence: Read the abstract and methods to decide whether the study answers your question.
- Network: Follow references and later citations along both supporting and opposing paths.
Create a source matrix
| Paper | Population | Method | Finding | Limitation | Relevance |
|---|---|---|---|---|---|
| One row each | More than topic | Experiment, survey or review | In your own words | What authors concede | Why include or exclude |
Finally inspect who is absent
Ask whether the results contain only one language, institution type, method or position. A complete search does not merely collect support; it actively seeks evidence that would make the conclusion more careful.
References
- WAME: Generative AI and Scholarly Manuscripts
- UNESCO: Guidance for Generative AI in Education and Research
Discover more from Geoffrey Chen
Subscribe to get the latest posts sent to your email.