[Research Project] Understanding Cognitive States From Motion

Understanding human cognitive states such as hesitation, uncertainty, and readiness is critical for improving virtual reality (VR) interactions. This study explores the potential of machine learning (ML) models to predict and interpret these states based on users’ motion data captured by VR headsets and controllers. Unlike traditional approaches that rely on subjective self-reports, our method quantifies hesitation using objective motion-based measures. We design a VR task that elicits hesitation and annotate hesitation levels as continuous values rather than binary labels. Using this dataset, we train ML models to predict hesitation and compare their performance against human judgment. Furthermore, we employ model visualization techniques to analyze which features the ML models prioritize when interpreting cognitive states. Our findings will provide insights into decision-making processes in VR, contributing to more adaptive and responsive VR systems.