A social interaction field model accurately identifies static and dynamic social groupings

"Identifying people who gather in a crowd is an essential cognitive function of our social lives. To achieve that, our brain needs to identify the interactions between individuals and integrate them into groups. Previous studies have mainly focused on the perceptual grouping of low-level visual features. However, very little attention has been paid to grouping in a social scene. In this study, we implemented cutting-edge virtual reality technology whereby we could manipulate characteristics of avatars in virtual scenes in a quantitative way. We conducted sophisticated psychophysical experiments to investigate the effect of interpersonal distance and angle on the judgments of social groups. Based on that, we deduced a space where people interact with each other and developed a social grouping field model (SGFM). The model was further verified by an experiment using multiple avatars in complex virtual spaces that could well predict human participants’ judgments in real social scenes that the conceptualized proxemic theory and conventional perceptual association grouping model failed to achieve. Our findings indicate an efficient computational framework in analyzing social interactions, which provides insight into how our brain organizes and interprets social information, enabling identification of social groups.