Neuroadaptive Haptics: A Proof-of-concept Comparing Reinforcement Learning from Explicit Ratings and Neural Signals for Adaptive XR Systems

Lukas Gehrke^1*Aleksandrs Koselevs¹Marius Klug²Klaus Gramann¹

• ¹Technical University of Berlin, Berlin, Germany
• ²Brandenburg University of Technology Cottbus-Senftenberg, Senftenberg, Brandenburg, Germany

Neuroadaptive haptics offers a promising path to immersive extended reality (XR) by dynamically tuning multisensory feedback to user preferences. We present a system that adapts XR haptic rendering through reinforcement learning (RL), using either explicit user ratings or implicit signals decoded from Electroencephalography (EEG). In a user study, participants interacted with virtual objects in VR while EEG data were recorded and decoded to infer user experience. Our neural decoder achieved a mean F1 score of 0.8, supporting informative but noisy classification. In two RL conditions, haptic parameters were adapted based on either explicit or implicit rewards, with exploratory findings indicating unstable agent behavior. A limited number of interaction steps likely constrained exploration and contributed to convergence instability. Revisiting interaction design to support more frequent sampling may improve robustness to EEG noise and rating drift. By demonstrating RL-based adaptation from implicit neural signals, our proof-of-concept is a step towards seamless, low-friction personalization in XR.