EdTech & Innovation
Medical Students Accept LLM-Powered 3D Avatars for Shared Decision-Making Training
Bostraeten et alJMIR formative research

Shared decision-making (SDM) is a foundational competency in patient-centered care, but traditional simulation-based education using standardized patients is constrained by high costs, limited scalability, and inconsistent feedback quality. To address these barriers, investigators at KU Leuven developed a web-based educational platform that combines large language models (LLMs) with interactive 3D virtual patient avatars to provide real-time SDM communication training for undergraduate medical education (UME).
The platform was built through an iterative, interdisciplinary process involving medical educators, SDM experts, software developers, user experience designers, and AI engineers. The resulting system incorporates real-time spoken dialogue with diverse, photorealistic 3D avatars, customizable layered prompts that govern avatar personality and clinical scenarios, an educator case-authoring panel, and automated formative feedback mapped to 6 distinct phases of SDM.
A descriptive, cross-sectional mixed methods study evaluated student acceptance following 1 month of system access between November and December 2025. Among 25 recruited medical students, 24 completed a 9-item, 5-point Likert questionnaire based on the technology acceptance model (TAM). Internal consistency was acceptable for perceived usefulness (α=0.76) and intention to use (α=0.78), but did not reach the acceptable threshold for perceived ease of use (α=0.69). All construct means exceeded the midpoint of 3, with perceived usefulness scoring highest (mean 4.43, SD 0.50), intention to use scoring 4.42 (SD 0.59), and perceived ease of use scoring lowest (mean 3.81, SD 0.68).
Narrative analysis of responses from 14 students who provided optional text feedback indicated that learners appreciated the psychologically safe environment to rehearse complex communication skills. However, students noted reservations regarding behavioral realism compared to physical appearance and reported recurring technical issues, including microphone errors, application freezing, and crashes. The authors conclude that LLM-driven 3D avatars represent a feasible, well-accepted modality for SDM instruction, noting that future research must prioritize technical reliability and evaluate objective clinical learning outcomes in multi-institutional trials.
Key Takeaways
- In an acceptance study of 24 undergraduate medical students, the LLM-powered avatar platform achieved high acceptance above the scale midpoint of 3, with perceived usefulness rated at a mean of 4.43 (SD 0.50), intention to use at 4.42 (SD 0.59), and perceived ease of use at 3.81 (SD 0.68).
- Qualitative analysis revealed that participants valued the low-pressure practice environment free from peer or patient presence, but they identified technical malfunctions—such as microphone issues, freezing, and crashes—and behavioral realism as the primary areas needing improvement.
Read the original sourceJMIR formative research
