AI Virtual Patients Help Medical Students Practice Communication Skills

Researchers released a new dataset of 240 French doctor-patient conversations for training AI virtual patients. These virtual patients use large language models to simulate realistic clinical scenarios, helping medical students practice their communication skills without relying on human actors. The system allows trainers to control the patient's responses, making it a flexible tool for various training situations.

This development could improve medical education by providing consistent, accessible training for students. Currently, medical schools rely on human standardized patients, which are expensive and limited in availability. AI virtual patients could offer a cost-effective alternative, allowing students to practice more frequently and in diverse scenarios.

The dataset and pipeline are described in a paper on ArXiv. The research aims to address the gap in training availability by building a controllable LLM-based system for generating realistic patient interactions.