Proxemics for Human-Agent Interaction in Augmented Reality

Title of the Paper

Proxemics for Human-Agent Interaction in Augmented Reality

Paper Information

• Subject Area: Human-computer interaction and social spatial relationships in Augmented Reality (AR)
• Keywords: Proxemics, Augmented Reality, Personal Space, Virtual Agents, Human-Computer Interaction, Perception

Research Background and Problem

• Problem or Challenge: Existing AR design guidelines often overlook the personal space (PS) of virtual agents, which may significantly impact user behavior and physiological arousal.
• Importance: The social presence of virtual agents in AR environments can influence user behavior and comfort. Understanding the personal space of agents is crucial for optimizing human-agent interaction and designing comfortable AR experiences.
• Research Motivation:
  • Previous research in Virtual Reality (VR) has shown that people maintain a certain distance from virtual agents, and violations of this distance can lead to physiological arousal.
  • This study aims to explore whether users respect the personal space of virtual agents in AR and whether violations of personal space affect users' physiological arousal.
• Research Questions:
  • RQ1: Do users in AR respect the personal space of virtual agents, similar to maintaining a distance (>1 m) from strangers in real life?
  • RQ2: Does violating the personal space of virtual agents increase participants' physiological arousal?

Solution

• Proposed Method: Design and conduct an experiment involving 54 participants interacting with virtual agents in an AR art gallery scenario to evaluate user-agent distance preferences, behaviors, and physiological responses.
• Innovations:
  • Applying human interpersonal distance theory (Proxemics) to interactions with virtual agents in AR environments.
  • Comparing different types of virtual agents, including humanoid agents, robots, and cylindrical objects, to assess the impact of social characteristics on user behavior.
  • Incorporating physiological data (Electrodermal Activity, EDA) to quantify users' physiological arousal.
• Implementation Steps:
  1. Generate six virtual agents (two male, two female, one robot, and one cylindrical object) using an AR platform (e.g., Microsoft HoloLens 2).
  2. Design two experimental scenarios:
     • Block 1: Users approach and ask for directions, observing the "stopping distance" (preferred distance).
     • Block 2: Users walk through the agent, assessing changes in skin conductance.
  3. Collect behavioral data (distance) and physiological data (EDA), supplemented by subjective perception surveys.
  4. Analyze data using ANOVA, skin conductance signal processing, and qualitative feedback analysis.

Research Findings

• Specific Findings:
  • Users do respect the personal space of virtual agents in AR, maintaining an average distance of approximately 1 m.
  • Significant distance differences were observed between agent types: for example, male agents and cylindrical objects elicited greater distances, while female agents elicited shorter distances.
  • Skin conductance activity significantly increased when users walked through virtual agents, indicating heightened physiological arousal.
• Advantages Over Existing Solutions:
  • Provides clear social spatial standards for designing virtual agents in AR (e.g., personal space size based on gender or characteristics).
  • Emphasizes the importance of balancing immersion and social spatial comfort in AR environments.
• Experimental or Evaluation Results:
  • Physiological data analysis (EDA) revealed that walking through virtual agents, especially humanoid agents, increased users' physiological responses, reflecting potential negative effects of unnatural or overly close interactions.
  • Users preferred agents with social characteristics but also reported discomfort when passing through these agents.
• Limitations and Future Directions:
  • Limitations: Technical issues in the experimental environment, limited richness of virtual agent animations, and lack of diversity in participant demographics (race and gender).
  • Future Directions:
    • Investigate how the dynamic behaviors of virtual agents (e.g., movement, orientation, emotional expressions) influence human-agent interaction distances.
    • Explore whether the social characteristics of virtual agents (e.g., race or diversity) affect the allocation of personal space.
    • Further study the interaction of PS: whether there is dynamic adjustment or conflict between users' personal space and that of virtual agents.