ASafePlace: User-Led Personalization of VR Relaxation via an Art Therapy Activity

Paper Title

ASafePlace: User-Led Personalization of VR Relaxation via an Art Therapy Activity

Publication Info

• Topic area: Personalized VR biofeedback systems for relaxation training
• Keywords: VR relaxation, biofeedback, personalization, art therapy, Safe Place technique, generative AI, user-authored environments, autobiographical memory, physiological relaxation, immersive technology

Background and Problem

• Problem / challenge: Existing VR biofeedback systems for relaxation training lack deep personalization, relying on superficial configuration options or pre-built scenarios. This limits their ability to engage users and address individual emotional and sensory needs.
• Significance: Personalization based on lived experiences and memories can enhance the emotional resonance, engagement, and effectiveness of relaxation interventions, but such approaches remain underexplored in VR biofeedback systems.
• Motivation and related work: Prior studies have highlighted the importance of personalization in relaxation tools and the potential of art therapy for eliciting meaningful personal expressions. However, few systems integrate art-therapy-inspired activities to create deeply personalized VR environments and relaxation guidance.

Solution

• Proposed approach: ASafePlace, a VR biofeedback relaxation system that uses an AI-powered, art-therapy-inspired activity called The Safe Place to create personalized 360° virtual environments and audio guidance based on users’ autobiographical memories and sensory preferences.
• Novelty:
  1. Integration of art-therapy-inspired activities to generate deeply personalized VR environments and relaxation guidance.
  2. Use of generative AI and LLMs to transform user sketches and verbal descriptions into immersive 360° environments and personalized audio scripts.
  3. Demonstration of the synergistic effects of personalization and biofeedback on physiological and subjective relaxation outcomes.
• Procedure and key techniques:
  1. Personalization Phase: Users engage in a Safe Place exercise, imagining and sketching a personal sanctuary, and describing sensory details and memories in a therapeutic dialogue. Generative AI transforms these inputs into a 360° environment and personalized audio guidance.
  2. Relaxation Phase: Users perform breathing-based relaxation training while immersed in their personalized VR environment, with biofeedback visualized through particle effects and accompanied by personalized audio guidance.

Results

• Concrete findings:
  • HRV-SDNN increased significantly in all groups during relaxation, with the largest increase in the personalization-with-biofeedback (PB) group (+42.5%).
  • Respiration rate decreased most significantly in the PB group (-45.2%).
  • Subjective relaxation (RRS) and anxiety reduction (STAI-S) improved across all groups, with the PB group showing the largest but not statistically significant differences.
  • The PB group reported the highest sense of presence (IPQ overall score: 0.84) and perceived relevance (5.77/7).
• Advantage over baselines:
  • Personalization combined with biofeedback (PB) outperformed other conditions in physiological relaxation (HRV, respiration rate) and subjective presence and relevance.
  • Personalization alone (PN) enhanced involvement and relevance compared to non-personalized conditions.
• Experiments / evaluation:
  • 2×2 factorial design with 52 participants across four conditions: PB, PN, NB (biofeedback only), NN (neither personalization nor biofeedback).
  • Measurements included physiological data (HRV, respiration rate), subjective scales (RRS, STAI-S, IPQ, FSS), and qualitative interviews.
• Limitations and future work:
  • Limited sensory modalities (no haptic feedback or background sounds).
  • Semi-manual workflow requiring therapist facilitation.
  • Cultural and individual differences in Safe Place preferences.
  • Future work could explore automated workflows, multi-modal interfaces, and culturally adaptive designs.

Summary

ASafePlace is a VR biofeedback relaxation system that integrates art-therapy-inspired activities and generative AI to create deeply personalized 360° environments and audio guidance. Quantitative results demonstrated that personalization and biofeedback synergistically enhanced physiological and subjective relaxation outcomes, with the personalization-with-biofeedback condition showing the greatest improvements. Qualitative findings highlighted the importance of memory-driven personalization and sensory cues in fostering user engagement and emotional resonance. This work offers design implications for creating meaningful, personalized relaxation experiences and suggests broader applications in wellbeing and therapeutic contexts.