CoEmpaTeam: Enhancing Cognitive Empathy using LLM-based Avatars and Dynamic Role Play in Virtual Reality

Abstract

Cognitive empathy, the ability to understand others' perspectives, is essential for effective communication, reducing biases, and constructive negotiation. However, this skill is declining in a performance-driven society, which prioritizes efficiency over perspective-taking. Here, the training of cognitive empathy is challenging because it is a subtle, hard-to-perceive soft skill. To address this, we developed CoEmpaTeam, a VR-based system that enables users to train their cognitive empathy by using LLM-driven avatars with different personalities. Through dynamic role play, users actively engage in perspective-taking, experiencing situations through another person's eyes. CoEmpaTeam deploys three avatars who significantly differ in their personality, validated by a technical evaluation and an online experiment (n=90). Next, we evaluated the system through a lab experiment with 32 participants who performed three sessions across two weeks, followed by a one-week diary study. Our results showed a significant increase in cognitive empathy, which, according to participants, transferred into their real lives.

Figures (7)

• Figure 1: Avatars used in the CoEmpaTeam system. From left to right: Alice, Benji, and Caden.
• Figure 2: CoEmpaTeam system interface and role-play workflow: (a) Welcome screen for avatar selection. (b) Role-specific information with options to start the task or finish speaking. (c) Shared house rules organized into five categories. (d) Example of basic information for Alice to support role-play. (e-f) Motion-synchronized avatar arms when embodying Benji and Caden (Alice analogous).
• Figure 3: Example frames from the evaluation videos. Each condition featured a different target avatar placed in the center of the scene: (a) Alice, (b) Benji, and (c) Caden. The other two avatars were positioned to the sides to simulate a realistic group interaction while directing participants’ attention to the evaluated role.
• Figure 4: Pre- and post-training changes in IRI subscales. Both PT and FS increased significantly.
• Figure 5: Session-wise IRI scores. PT and FS were elevated relative to pre-training; PT showed session-to-session fluctuations, whereas FS showed no significant differences among the three training sessions.