Roleplay-doh
Enabling domain-experts to create LLM-simulated patients via eliciting and adhering to principles
Abstract
Creating realistic simulated patients for medical training is challenging and expensive. We introduce Roleplay-doh, a framework that enables domain experts to create LLM-simulated patients by eliciting and adhering to principles. Our approach allows medical professionals to define patient characteristics and behaviors through natural language principles, which are then used to guide LLM responses. This democratizes the creation of high-quality simulated patients while maintaining clinical accuracy and educational value.