Audio-Only Versus Video for Social Acceptability

Determine whether always-on audio capture on smart glasses mitigates social acceptability concerns compared to continuous video capture in the context of VisionClaw-style wearable AI agents.

Background

The paper discusses privacy and social acceptability issues arising from continuous egocentric sensing coupled with autonomous action. While the authors note that many useful VisionClaw interactions can operate without video (e.g., email briefings, calendar checks), they emphasize uncertainty about whether relying on audio-only capture improves bystander acceptance.

This problem is central to deploying always-on wearable agents in public and social settings, where perceptions of recording—especially when tied to autonomous identification or action—can significantly impact adoption and ethical use.