“How can AI prepare future doctors for difficult conversations with patients?”

Talking to patients cannot be learned from theory alone. Students have to take medical histories, convey sensitive or distressing information and learn to respond to very different reactions from patients, emotional ones included. Human simulated patients are especially valuable for this, but they are in limited supply.

In the AIMED research project (Artificial Intelligence in Medical Education), Zeitgeist Labs and Charité are running a randomised controlled trial. It asks whether, and how, simulated patients played by a voice AI help students learn to talk to patients. Zeitgeist Labs develops both the systems themselves and the technical research infrastructure the study runs on. Each simulated patient comes with a defined biography, a clinical situation, a level of knowledge about their illness and their own emotional reactions. The simulated patients react naturally to the way each student leads the conversation, and still keep reliably to their clinical case.

Students can practise difficult patient conversations on their own, again and again. The important step comes afterwards: the study tests what this new kind of training actually achieves. It examines whether, and under which conditions, a new technology becomes an effective method for medical education.