Interactive patient history
A simulated patient to practise taking a history with: the student is the doctor, the bot is Charlie, and only at the end does it show whether the right questions were asked.
What it is: a role play. The student takes the part of the doctor and talks to Charlie to work out what is wrong. The bot stays in character; the conversation ends with the question of what the next step would be.
Made by: Erik Elings, Amsterdam UMC. Runs on: edudatabase.ctl-vu.nl, not on this site.
What the Education Lab is testing here
Whether generative AI can supply believable counterparts for skills that normally cost a lot of people’s time. The learning is not in the bot’s answer but in practising the order of questions, precision and clinical reasoning. This is the app side of the programme: not only assistants for staff but interactive learning environments for students.
Why this works
Practising a patient history needs a counterpart who does not help. A fellow student who knows the case gives away what is going on without meaning to. A simulated patient who answers only what is asked lets the student notice when a question was never put - which is exactly the mistake you want them to be able to make without consequences.
What you need
- No preparation from the student. The conversation starts by saying hello.
- A debrief afterwards. The bot does not assess, it plays. The learning is in what the student does with the conversation next.
What to watch
This is an exercise, not an assessment, and certainly not a diagnostic instrument. Agree that before students start. And no real patient data goes in, anonymised or not.
Building something like it
The shape travels well beyond medicine: a difficult conversation with a parent, a client with an unclear brief, a colleague who did not keep an agreement. What makes it work is that the bot stays in role and does not turn helpful.
Try the patient history on edudatabase.ctl-vu.nl