Abstract

Virtual reality exposure therapy (VRET) has emerged as a promising treatment for specific phobias. This study introduces an AI-enhanced VRET system that dynamically adjusts exposure intensity based on real-time physiological feedback including heart rate variability, galvanic skin response, and facial expression analysis. The system employs a reinforcement learning algorithm to optimize the pace and intensity of exposure scenarios for each patient. A randomized controlled trial with 120 participants diagnosed with acrophobia (fear of heights) compared AI-VRET against standard VRET and traditional cognitive behavioral therapy. Results showed that AI-VRET produced significantly greater reductions in phobia severity (BAT scores improved by 68% vs. 52% for standard VRET and 41% for CBT) and required fewer sessions (mean = 6.2 vs. 9.1 and 12.3 respectively). The AI system demonstrated the ability to identify optimal therapeutic windows and prevent both under-exposure and overwhelming anxiety responses.