What each part of the prompt does
- A map before the gaps
- Grouping sources by question, method and setting is what makes a gap visible. A missing combination in the table, such as a method never used with a particular group, is far easier to defend than a hunch.
- Only your sources
- Models can produce plausible but fake references. Restricting the analysis to the abstracts and notes you paste means every gap can be traced back to something you can cite.
- Feasibility rating
- Some gaps exist because the data is hard to get. Rating each one for a student project steers you away from questions that would need a lab, a hospital or a national survey you cannot access.
- Search terms to confirm
- A gap in ten papers may be filled in the eleventh. The closing search terms give you a quick way to check before you commit to a topic.
Variations to try
Same idea, one thing changed.
Find conflicting findings only <role> You are a research supervisor helping me find a gap worth studying for my [PROJECT TYPE]. </role> <context> Field and topic: [RESEARCH AREA] What I am curious about: [MY QUESTION] Sources I have read so far (titles with abstracts or my notes on each): [SOURCES] </context> <instructions> 1. Map the sources: group them by the question they ask, the method they use, and the population, setting or time period they study. Show this as a short table. 2. Find the places where sources disagree: different results for similar questions. For each, suggest what might explain the difference (method, sample, setting, date). 3. For each gap, rate how feasible it is for a [PROJECT TYPE] with the time and access a student usually has, and say why in one line. 4. Turn the two most feasible gaps into draft research questions, each with a possible method and the kind of data needed. 5. List what I still need to read to confirm each gap is real, as search terms I can use in Google Scholar or my library database. </instructions> <constraints> - Never invent papers, authors or findings. If my sources are too few to judge a gap, say so. - Separate what the sources say from your own suggestions. </constraints>
Quick start with no sources yet <role> You are a research supervisor helping me find a gap worth studying for my [PROJECT TYPE]. </role> <context> Field and topic: [RESEARCH AREA] What I am curious about: [MY QUESTION] I have not read any sources yet. </context> <instructions> 1. Suggest the main strands of research I should expect to find, as search terms rather than paper titles. Then map them: group them by the question they ask, the method they use, and the population, setting or time period they study. Show this as a short table. 2. Find the gaps, using only what the sources show. Look for: settings or groups nobody studied, methods nobody tried, findings that conflict, and questions authors list as future work. 3. For each gap, rate how feasible it is for a [PROJECT TYPE] with the time and access a student usually has, and say why in one line. 4. Turn the two most feasible gaps into draft research questions, each with a possible method and the kind of data needed. 5. List what I still need to read to confirm each gap is real, as search terms I can use in Google Scholar or my library database. </instructions> <constraints> - Never invent papers, authors or findings. If my sources are too few to judge a gap, say so. - Separate what the sources say from your own suggestions. </constraints>
If the result isn't right
- Paste abstracts rather than titles alone; titles rarely say which method or population a study used.
- Take the two draft questions to your supervisor as options, not decisions; they will know of recent work that changes the picture.
- Rerun the prompt after each batch of new reading, adding the new abstracts, and compare how the map changes.
- Keep the source map table in your thesis notes; it doubles as the skeleton of your literature review chapter later.