MAAS Practice: an AI patient for clinical reasoning
An AI patient for clinical reasoning, steered by published MAAS sources and checked before students use it.
Research barely looks at early practice: a 2025 review of 50 studies on teaching clinical reasoning found three that looked at year 1. In MAAS Practice, students practise clinical reasoning with an AI patient, calmly and as often as they need, from year 1. The student asks, listens and links each answer to the underlying mechanism. The patient gives each fact only when the student asks for it, so every question shows what the student is looking for. After each case the feedback checks the telling question, the link to the mechanism, the reasoning put into words and the danger kept open, and gives one next step.
We present MAAS Practice at the symposium “From AI literacy to critical use of generative AI: theory and action” of Amsterdam UMC, the NVMO working group Digital Learning and Innovation and VOR ICT in Education, on 16 November 2026.
We built the AI patient in two steps: steer, then check. Published MAAS sources steer it: the Review of Systems shapes the case, the Medical Interview handbook the patient and the Clinical Reasoning handbook the feedback. AI test students then work each case: a strong student, an interrogator and an early closer. Students get a case only once it passes all 48 checks.
The MAAS Case Complexity Code orders cases by the number of possible mechanisms, the nature of the problem and the burdening contexts, so teachers can choose the level that fits their curriculum. The first three cases follow one complaint, a burning feeling behind the breastbone, at three levels: Sarah with reflux (1.0.0), Sarah two weeks later with unmentioned ibuprofen (1.0.1), and Rob with angina he plays down (1.1.1).
These three cases are a proof of concept, and we build the next cases with teachers. Try a case, judge the feedback and tell us what works: info@maas-mi.eu.
Open: MAAS Practice
