Wait for It: A Component-Level Analysis of Latency Feedback Mechanisms for LLM Agents in VR
TL;DR
It is suggested that natural, behavior-based feedback maintains the illusion of conversing with a thinking entity, while technical indicators expose the computational substrate in AI agents.
Abstract
Delays in AI responses pose a significant challenge to immersive Virtual Reality experiences. While previous research demonstrates that feedback can reduce the perception of waiting, we lack understanding of which types work best and why. We conducted a within-subjects study with 18 participants to systematically compare three isolated feedback approaches during VR agent interactions: verbal fillers (e.g., um..., hmm...), visual indicators (loading spinners), and embodied gestures (thinking animations). Our results revealed that embodied gestures significantly outperformed visual loading indicators across multiple dimensions, making agents feel more lifelike and socially present. Interestingly, explicit visual feedback like loading spinners actually degraded perceived humanness and broke immersion by forcing users to shift from social to technical mental models. These findings suggest that natural, behavior-based feedback maintains the illusion of conversing with a thinking entity, while technical indicators expose the computational substrate. We present empirically-grounded design guidelines for creating more engaging AI-powered VR agents.
