Estimating Detection Thresholds of Being Looked at in Virtual Reality for Avatar Redirection

Research Background and Problem Statement

• What issues or challenges did the authors identify?
  In virtual reality (VR), a critical nonverbal cue in social interactions—gaze direction—is often simplified or overlooked. Particularly in multi-user collaborative virtual environments, misaligned gaze directions can negatively impact users' sense of social presence and interaction quality. The authors point out that previous research on gaze perception has primarily focused on the physical world, while studies in virtual environments remain scarce.

• Why is this issue important?
  Gaze direction plays a crucial role in conversations and collaborations, influencing attention allocation, interaction roles, and feelings of intimacy. In virtual social environments, accurately perceiving whether someone is looking at you is fundamental to enhancing social presence and user experience. Additionally, this issue is closely related to spatial desynchronization in VR, directly affecting the effectiveness and consistency of multi-user interactions.

• Research Motivation and Related Work
  Inspired by the demands of social scenarios in virtual reality applications and building on research findings from gaze perception in the physical world, the authors designed a series of experiments to investigate users' detection thresholds for perceiving gaze in virtual environments. They also explored whether these thresholds align with those in the physical world. Furthermore, the study examined the impact of gaze misalignment on social presence, providing a foundation for future avatar redirection technologies.

Proposed Solution

• What methods or solutions did the authors propose?
  The authors designed two experiments: one focused on the impact of avatar gaze misalignment on social presence, and the other employed a two-alternative forced-choice (2AFC) experimental method to quantify the detection thresholds for gaze perception. The study examined the effects of gaze deviation angle, avatar distance, gaze direction (left/right), and stimulus duration on perception.

• What is innovative about this solution?

  1. First-ever systematic investigation of gaze perception thresholds in VR, building on research from the physical world to explore differences in virtual environments.
  2. Combined real-world interaction scenarios (e.g., the "Guess Who I Am" game) with controlled experimental tasks to comprehensively analyze the relationship between subjective social impact of gaze misalignment and objective perception accuracy.
  3. Provided specific gaze perception thresholds for different distances and durations, clarifying subtle differences between gaze perception in virtual and physical environments.

• What are the implementation steps and key technologies used?

  1. Experiment 1: Analysis of Social Presence
     • A virtual museum scene was used, where a virtual avatar interacted with the user in a "Guess Who I Am" game to simulate social interaction. The avatar's gaze direction was manipulated under three conditions (direct gaze, slight deviation, and significant deviation).
     • The Networked Minds Social Presence Inventory was used to evaluate user experience differences under different conditions.
  2. Experiment 2: Evaluation of Gaze Perception Thresholds
     • A 2AFC staircase procedure was employed, gradually increasing or decreasing the avatar's head deviation angle until the user reported no longer feeling gazed at.
     • Three variables were independently manipulated: gaze direction (left/right), avatar distance (1.1 meters and 3.4 meters), and stimulus duration (300 ms and unlimited).

Research Findings

• What specific findings were obtained?

  1. The study identified precise thresholds for users' perception of avatar gaze in virtual environments: approximately 5.035° for short-duration stimuli and 6.590° for long-duration stimuli.
  2. Social presence was significantly influenced by whether users noticed the avatar's gaze deviation. When the deviation was noticed, social presence decreased significantly.
  3. Stimulus duration had the greatest impact on perception accuracy, with longer stimuli significantly increasing perception thresholds, while distance and gaze direction had minimal or insignificant effects.

• What advantages does it have compared to existing solutions?

  • The research fills a gap in the study of gaze direction perception in virtual environments, providing a systematic approach to experimental design and data analysis.
  • The findings offer operational guidelines for future behavioral models of VR avatars and virtual agents, particularly for applications aimed at enhancing social presence.

• What were the experimental or evaluation results?

  • Quantitative experiments demonstrated that perception thresholds remain stable within a certain range but are significantly influenced by stimulus duration and context.
  • In real-time social interaction tasks, users' detection rates for significant deviations (13°) were lower than expected, suggesting that perception thresholds might expand in real interactions due to the richness of nonverbal information.

• Limitations and Future Directions

  1. Limitations
     • All experiments were based on virtual avatar agents (non-human), so the applicability of the results to human interactions remains uncertain.
     • The avatar styles and complexity were limited, and the study did not explore how more realistic or cartoonish avatars might affect perception.
     • Significant individual differences were observed, with some participants exhibiting large perception deviations. Further research is needed to explore influencing factors such as visual habits or cultural background.
  2. Future Directions
     • Extend the study to more realistic human avatars or multi-user interaction scenarios.
     • Investigate potential causes of individual differences among groups, such as gaming experience or perceptual styles.
     • Apply the study's findings to VR multi-user collaboration, particularly in addressing avatar redirection challenges in spatial desynchronization.

Through this research, the authors provide experimental evidence for design decisions aimed at enhancing user experience in future VR social applications, laying a theoretical foundation for the development of more natural and realistic virtual interaction experiences.