The Role of Personality of Conversational Virtual Avatars on Proxemic Behaviour during Indoor Navigation

Paper Title

The Role of Personality of Conversational Virtual Avatars on Proxemic Behaviour during Indoor Navigation

Publication Info

• Topic area: Human-avatar proxemics in augmented reality indoor navigation.
• Keywords: Proxemics, conversational virtual avatars, augmented reality, personality traits, indoor navigation, interpersonal distance, walking speed, human-computer interaction.

Background and Problem

• Problem / challenge: Existing research on proxemics in human-avatar interactions focuses primarily on static or one-sided scenarios, neglecting dynamic navigation where both the user and avatar are in motion. Additionally, the influence of avatar personality traits expressed through speech on proxemic behavior remains underexplored.
• Significance: Understanding proxemic behavior in dynamic navigation scenarios is critical for designing effective and socially competent avatars for real-world applications, such as museums, hospitals, and industrial environments.
• Motivation and related work: Prior studies have shown that personality traits influence interpersonal distances in human-human and human-avatar interactions. However, these studies have not addressed how personality traits, expressed verbally, interact with motion-related factors like walking speed to shape proxemic behavior during navigation.

Solution

• Proposed approach: An AR indoor navigation system featuring a Conversational Virtual Avatar (CVA) with three distinct personalities—Warm, Dominant, and Conscientious—guiding users to destinations at three walking speeds (Low, Medium, High).
• Novelty:
  1. Investigates dynamic proxemic behavior during navigation, where both the user and avatar are in motion.
  2. Explores the influence of avatar personality traits, expressed through speech, on interpersonal distance.
  3. Examines the interaction between avatar walking speed and personality on proxemic behavior.
  4. Provides design guidelines for avatar personality and motion in AR-based navigation systems.
• Procedure and key techniques:
  • Developed an AR navigation system using Meta Quest 3 HMD and the ConvAI API to generate avatars with controllable personality traits.
  • Conducted a between-subjects study (N=27) with three personality conditions (Warm, Dominant, Conscientious) and three walking speeds (Low, Medium, High).
  • Measured static and dynamic interpersonal distances (IPD) and assessed subjective workload using the NASA-TLX questionnaire.

Results

• Concrete findings:
  • Static IPD: Dominant avatars elicited the highest median IPD (1.89 m), followed by Conscientious (1.86 m) and Warm (1.83 m).
  • Dynamic IPD: Warm avatars elicited the largest median IPD (1.97 m), followed by Conscientious (1.91 m) and Dominant (1.88 m).
  • Faster walking speeds led to greater dynamic IPD, with High speed (1.2×) resulting in the largest distances.
  • Conscientious avatars were associated with the lowest cognitive workload (NASA-TLX score: 3.24), while Dominant avatars had the highest workload (5.67).
• Advantage over baselines: Demonstrates how personality traits and walking speed jointly influence proxemic behavior, extending prior static-focused studies to dynamic navigation scenarios.
• Experiments / evaluation:
  • Mixed-method study with 27 participants, each assigned to one of three personality conditions.
  • Measured static and dynamic IPD, preferred walking speed, and subjective workload.
  • Used non-parametric statistical tests (Wilcoxon rank-sum) to analyze IPD differences.
• Limitations and future work:
  • Small sample size (N=27) and limited cultural diversity reduce generalizability.
  • Personality was expressed only through speech; future work should explore non-verbal cues like gestures and prosody.
  • Dominant avatars were less effective in this context but may be suitable for high-urgency scenarios like emergency response.

Summary

This study investigates how the personality and walking speed of a Conversational Virtual Avatar (CVA) influence human-avatar proxemic behavior during indoor navigation in augmented reality. Results show that avatar personality significantly affects interpersonal distance (IPD), cognitive workload, and preferred walking speed, with Conscientious avatars being the most cognitively efficient and Warm avatars eliciting the largest dynamic IPD. Faster walking speeds consistently increased IPD. These findings provide actionable design guidelines for tailoring avatar personality and motion to specific contexts, such as task-focused environments (e.g., hospitals) or socially engaging settings (e.g., museums). Future work will explore non-verbal personality cues and high-urgency scenarios.