This paper introduces a multi-agent SDM simulation framework leveraging generative agents to simulate realistic patient-doctor interactions for healthcare communication research. The framework includes Patient, Doctor, and Measurement agents, enabling systematic evaluation of SDM interventions and automated measurement using validated instruments.
Key findings
The proposed framework enables scalable realism, systematic variation, and automated evaluation in SDM research.
Generative agents can simulate diverse patient profiles and exhibit emergent social behaviors mirroring human dynamics.
The framework provides immediate feedback at scale and addresses critical gaps in SDM research.
Limitations & open questions
The paper identifies key risks including model hallucinations, representational biases, and ecological validity.