Facing Yourself: A Proxemic Approach to Self-Dyadic Interactions in VR

Abstract

Virtual Reality (VR) provides a powerful framework for studying human social interactions, yet generating realistic and diverse interaction data remains a major challenge. Asynchronous Interaction (AsI) paradigms offer a promising alternative by enabling users to interact with their own previously recorded behaviours. However, their ability to reproduce fine-grained social mechanisms is still unclear. To investigate whether AsI preserves proxemic behaviours observed in dyadic interactions, we carried out two user studies. The first study leverages AsI and VR to capture social dyadic interactions, where participants alternated between approaching a virtual human and being approached by a virtual human animated from their prior recordings. We analyze AsI through four key factors: gender, approach orientation, movement velocity and attractiveness. Results show that asynchronous interactions reproduce several established proxemic effects: women maintain larger distances and exhibit greater reactive displacements, frontal approaches elicit stronger avoidance than lateral ones, and velocity modulates both displacement and gender effects. Moreover movements and agents perceived as less attractive tend to elicit stronger reactions. In the second study, we assessed the perceived realism, behaviour, awareness and appropriateness of interaction distances using videos of both asynchronous and synchronous interactions (the latter created through a traditional synchronous approach). Our findings support the fact that AsI preserves several measurable proxemic regularities, and confirm the use of AsI as a valid and scalable framework for studying social behaviour in VR and for generating interaction datasets. Finally, we provide guidelines for designing socially aware virtual agents.