Vibrotactile Preference Learning: Uncertainty-Aware Preference Learning for Personalized Vibration Feedback

A research team from Stanford University and the University of Southern California has developed Vibrotactile Preference Learning (VPL), a novel AI system designed to personalize haptic feedback by learning individual user preferences. The system addresses a fundamental challenge in human-computer interaction: people perceive vibrations differently, making one-size-fits-all feedback ineffective. VPL employs Gaussian-process-based, uncertainty-aware preference learning to map each user's unique tactile preferences across multiple vibration parameters.

During interaction, VPL guides users through 40 rounds of pairwise comparisons where they choose between different vibration patterns, while also reporting their confidence in each choice. This data, combined with an expected information gain acquisition strategy, allows the system to efficiently explore the parameter space and build accurate preference models. In a user study with 13 participants using Microsoft Xbox controller feedback, VPL successfully learned individualized preferences while maintaining comfortable, low-workload interactions.

The research, accepted to the ACM UMAP 2024 conference, demonstrates that VPL can scale personalization of vibrotactile experiences—a capability becoming increasingly important as haptic feedback expands beyond gaming into VR, automotive interfaces, and accessibility technologies. The system's efficiency (requiring only 40 comparison rounds) and attention to user comfort make it practical for real-world deployment where users won't tolerate lengthy calibration processes.