Set stage-by-stage AI rules for your trainees
My residents use AI for notes and differentials, and I can't tell which uses help them learn and which let them skip the thinking.
Unvalidated framework No patient data Individual access available
Source check: 2026-11-11 · Physician review: not yet recorded
The workflow
- List the AI uses your students and residents already rely on, such as record summaries, note drafts, and differentials.
- Rank each by what the tool hands over: pulling up data is low, drafting a note is moderate, and supplying the differential and plan is high.
- For each training stage, mark each use as defer, introduce under supervision, or sustain, and write down who is accountable for AI output.
- Share the list at the start of the rotation, and where you have marked differentials as introduce, have learners commit to their own differential before using AI to check for missed diagnoses.
- Check unaided reasoning with oral exams, observed encounters, or written cases that show how the learner reached the answer.
Tools and cost
USD unless stated otherwise. Billing periods and eligibility matter. A tool listed here was not necessarily tested in the cited study.
What the evidence supports
The article proposes a stage-based framework for introducing AI in medical training. It is a consensus proposal, not a prospective validation of educational outcomes.
Published framework; not clinical validation · Read the source ↗
Where it goes wrong
Use it as a discussion framework and measure learners’ unaided reasoning. The framework itself does not show that a particular training policy prevents overreliance.