Yeyu Wang studies how clinical teams learn, coordinate, and make decisions during high-stakes medical simulations. Her research develops interpretable multimodal learning analytics that combine human insight with advanced computational methods. As medical education increasingly incorporates virtual reality simulations, instructors can observe not only communication but also gaze patterns, actions, and indicators of cognitive load. The challenge is making sense of these complex data in ways that remain transparent, trustworthy, and useful for training. Yeyu addresses this challenge by integrating two methodological pillars: Quantitative Ethnography, which unifies qualitative interpretation with statistical modeling, and multimodal machine learning, which detects patterns across diverse data streams. Her goal is to create analytic systems that reveal how collaborative expertise develops moment to moment as medical teams respond to dynamic emergencies. By modeling how communication, perception, and cognitive load interact within a team, her work contributes to theory building about how expertise is enacted under pressure.
As a MIDAS Fellow, Yeyu will extend these methods by incorporating large language models and multimodal machine learning to automate parts of the analytic pipeline while preserving interpretability. She will develop systems that flag uncertain model outputs for human review, adapt temporal models to capture how one modality influences another, and generate real-time or post-simulation feedback for instructors and learners. This work advances responsible, human-centered AI for clinical education, where transparency and accountability are essential.
Ultimately, Yeyu aims to help medical educators understand not only what teams do in simulation, but why they do it, and how their learning can be supported more effectively. Her research bridges AI, learning sciences, and medical training to improve clinical teamwork, decision making, and patient experience.