Reading an AI paper
15 questions to answer for any paper or vendor claim about a clinical AI tool. Write a short answer for each; a blank answer is itself a finding. Background is in Track 3 · Clinical AI evaluation and safety.
Paper / tool:
Reviewed by / date:
- 1.What clinical problem?
- 2.What dataset?
- 3.What population?
- 4.What labels?
- 5.What model?
- 6.What comparator?
- 7.What validation?
- 8.What metrics?
- 9.What calibration?
- 10.External validation?
- 11.Bias?
- 12.Clinical utility?
- 13.Generalizability?
- 14.Conflicts of interest?
- 15.Can this actually change clinical care?