Modeling Subjective User Experience in Haptics Through Individualized Bayesian Models.

Abstract

In many areas of haptics, evaluation has prioritized objective task performance, with subjective aspects of user experience (UX), such as emotional response and user preference, often being treated as complementary measures. In this work we present a unified modeling framework for UX in haptic and physical human-robot interaction that simultaneously evaluates metrics of performance, emotional response, and preference as a function of system design parameters. Our framework uses Bayesian hierarchical modeling to capture both individual variability and population-level trends. We apply this approach to a line-following task conducted with a custom-built haptic knob platform in a study with 30 participants, in which we measure task performance together with subjective ratings of emotion and preference while manipulating damping, task difficulty, and feedback given to participants on their performance. Results show that while population-level models capture consistent effects on performance, emotion and preference did not show similarly consistent population-level trends. Instead, subjective responses varied substantially across users, and examining individual models helped reveal participant-specific patterns that were not apparent from population-level summaries alone. Together, these findings highlight the importance of individualized modeling in the design and evaluation of haptic systems that aim to quantitatively account for the full spectrum of user experience.