Avatar-mediated communication in collaborative virtual environments: A study on users’ attention allocation and perception of social interactions

Collaborative virtual environments (CVEs) facilitate avatar-mediated communication (AMC), where users interact using human-like virtual characters in shared virtual worlds, enhancing the attractiveness, attentiveness, and connectedness of remote social experiences and thus becoming extremely popular nowadays in various application domains such as education and healthcare. Understanding how different aspects of avatar behaviours influence various types of social interactions is crucial for improving the design of CVEs. Grounded in a theoretical framework based on avatar anthropomorphic realism, nonverbal social cues, eye-mind hypothesis, and interaction process analysis, this study investigates the impact of avatars’ gaze behaviours on users’ attention allocation and perceptions during AMC in CVEs. A two-arm randomised controlled trial (RCT) with 60 participants (29 males and 31 females) compared static gaze and natural gaze avatars during socio-emotional and task interactions. Three-dimensional eye-tracking data revealed distinct attention patterns across three primary nonverbal social cues: eye gaze, head orientation, and pointing gesture. Furthermore, avatars’ gaze type and interaction type were both found to significantly affect participants’ attention allocation; natural gaze behaviour and task interactions mitigated the general gaze-avoidance pattern observed in previous studies. However, avatars’ gaze type did not impact participants’ perceptions of social presence and anxiety. This research provides a nuanced understanding of attention allocation across nonverbal social cues during AMC and underscores the importance of avatars’ gaze and interaction types, highlighting important implications for the future design of CVE to enhance attention coordination and communication. Additionally, it calls for more comprehensive studies to explore avatars’ anthropomorphic realism and its effects on user perceptions and overall experience during AMC. © 2025 The Authors.