JMIR Publications

ABSTRACT

Simulation is effective for learning and improving basic and advanced communication skills across health professions. Disclosure and apology (D&A) following a medical error is a high acuity, low occurrence (halo) event that is critical to patient safety, trust and clinical care. Many learners receive little training or practice in this important skill. Simulation experiences can be limited by availability of curriculum time and simulation resources. Generative AI-enabled avatar-based standardized patients can provide individualized, scalable simulation for practice and learning. The University of Massachusetts Chan Medical School (UMass Chan) developed and deployed a platform for training in medical error disclosure and apology for undergraduate medical learners and interprofessional graduate trainees. This model was based on curriculum originally delivered using human SPs. Consistent rubrics, principles and cases were translated into a GenAI-supported system, the GenAI Conversational Avatar Practice System (GenAI CAPS) accessible via virtual reality headset (VR) and desktop environments. In this system, GenAI avatars assumed both patient and educator roles, providing interactive engagement and rubric-based feedback aligned with traditional SP training and responsibilities. Pedagogic authenticity was maintained by interweaving scripted didactic portrayed by non-generative avatars in facilitator and peer learner roles. The pilot study described and analyzed was conducted with 93, 4th year undergraduate medical students who engaged with either traditional SPs (n = 47) or GenAI SPs (n =46 ). Outcomes included post-simulation learner perceptions, self-assessed communication behaviors, and mid-simulation emotional reflections. Learner-reported outcomes were comparable between GenAI and traditional SP modalities, with high levels of perceived effectiveness in both groups. Within the GenAI condition, improvements across sequential simulations suggest rapid learner adaptation to the platform. Learners reported a range of emotional responses consistent with authentic clinical encounters. These findings demonstrate the feasibility and educational potential of GenAI-enabled avatar-based simulation as a scalable complement to traditional SP programs. Ongoing work is extending this approach to additional communication-focused use cases and refining system design to enhance usability, realism, and accessibility.