Abstract
Recent advancements in artificial intelligence (AI) have significantly expanded opportunities for AI-human communication.
Concurrently, virtual reality (VR) technologies enable interactions that simulate shared physical spaces. The convergence
of these technologies suggests imminent growth in AI-human interactions within VR environments. As AI dialogue capabilities
approach human-like proficiency, distinguishing between AI and human interlocutors in avatar-mediated VR communication
may become increasingly challenging. This raises fundamental questions regarding human perception of AI
and appropriate interaction paradigms. Our research addresses these questions by investigating AI-human communication
in VR environments, specifically focusing on personal space (PS) characteristics during such interactions. We conducted a
2×2 between-participants experiment (N = 76) manipulating two factors: avatar identity (perceived as AI or Human) and
conversation presence (Conversation or No Conversation). Importantly, we defined avatar identities based on participants’
perceptions rather than the actual identity of the avatar’s operator. In this experiment, all participants actually interacted
with an AI, irrespective of whether they were assigned to the ‘AI’ or ‘Human’ condition. They were led to believe they
were interacting with the entity corresponding to their assigned condition through experimental instructions and procedures.
Our findings suggest a potentially unconscious extension of human social norms to AI entities in immersive environments,
particularly concerning PS. This interpretation is supported by the result that no significant differences in PS were found
based on perceived avatar identities. This lack of difference suggests that humans may, at least in this context, apply similar
social-psychological frameworks to AI as they do to human interlocutors. Furthermore, participants maintained significantly
smaller PS in conditions involving active ‘Conversation’ compared to those with ‘No Conversation’. This result indicates
that active communication itself is a powerful factor in reducing interpersonal distance requirements, regardless of whether
the interlocutor is perceived as human or AI, with important implications for avatar-based spatial design in VR environments.