VRBubble: Enhancing Peripheral Awareness of Avatars for People with Visual Impairments in Social Virtual Reality
Tiger F. Ji, Brianna Cochran, Yuhang Zhao · 2022 · Proceedings of the 24th International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS 2022) · doi:10.1145/3517428.3544821
Summary
This paper introduces VRBubble, an audio-based interaction technique designed to make social virtual reality environments accessible to people with visual impairments (PVI). While social VR platforms like VRChat and Rec Room are growing rapidly as spaces for remote socialization and collaboration, they rely almost entirely on visual information to convey the presence, identity, and movement of other users's avatars. This creates a fundamental accessibility barrier for PVI, who cannot perceive surrounding avatar dynamics that sighted users take for granted through peripheral vision. The researchers grounded their design in Hall's proxemic theory, which defines social distances corresponding to different types of interpersonal interaction. VRBubble divides the virtual space around a user into three concentric zones: an Intimate Bubble (within 1 foot, signaling potential collisions), a Conversation Bubble (within 4 feet, suitable for direct interaction), and a Social Bubble (within 12 feet, where potential conversation partners exist). The system provides spatial audio feedback when avatars enter, exit, or move between these zones. Through an iterative user-centered design process involving a formative study with six PVI, the team developed three distinct audio feedback alternatives: earcons (brief abstract sounds with pitch and timbre variations), verbal notifications (spoken name and position announcements), and real-world sound effects (ambient crowd noise and footsteps). Critically, VRBubble allows users to customize which audio alternatives they receive for different bubbles, avatar types (friends vs. strangers), and social contexts.
Key findings
The evaluation with 12 legally blind participants compared VRBubble against a standard audio beacon baseline across navigation and conversation tasks. In navigation, VRBubble significantly reduced avatar amount estimation error rates compared to the baseline (VRBubble mean error: 0.239 vs. baseline: 0.465, p < .001), demonstrating enhanced peripheral awareness. However, participants navigated significantly slower with VRBubble (mean: 36.5 seconds vs. baseline: 31.3 seconds), partly because the richer information prompted exploratory behavior. VRBubble was more distracting in crowded environments with high avatar counts, suggesting an upper limit of about five avatars that PVI can peripherally process during attention-demanding tasks. For avatar identification, participants recalled names with a mean accuracy of 0.715 during navigation, though this dropped below 0.5 in crowded scenarios due to overlapping verbal notifications. All 12 participants valued the bubble-based distance concept and the customization capability. Participants preferred verbal notifications during navigation (less attention-demanding) but favored earcons and real-world sound effects during conversation (less distracting). Five participants changed their audio customization between navigation and conversation tasks, confirming that context-dependent preferences are important. Participants rated VRBubble higher than baseline for both effectiveness (3.92 vs. 2.67) and immersion (4.33 vs. 2.67) in navigation.
Relevance
This research is significant for accessibility practitioners because it addresses the emerging challenge of making immersive social platforms accessible as VR adoption grows. The proxemics-based design framework offers a reusable pattern for translating spatial social information into non-visual modalities, applicable beyond VR to any context where spatial awareness matters for PVI. The finding that users need customizable audio feedback rather than one-size-fits-all solutions reinforces broader accessibility principles about user agency and personalization. The study's discovery of cognitive limits in processing peripheral audio information (approximately five avatars maximum during conversation) provides practical design guidance for audio-based awareness systems. The work also demonstrates that desktop-based VR, rather than headset-based VR, may be a more practical entry point for accessible social VR since PVI cannot benefit from headset visual displays.
Tags: virtual reality accessibility · visual impairment · spatial audio · proxemics · social interaction · peripheral awareness · audio feedback · earcons · assistive technology · avatar awareness