Agent Clusters: The Usual vs. The Unusual

Conversational clusters refer to groups of two or more people engaged in face-to-face interaction, wherein the arrangement of people’s bodies in space determines the spatial layout of clusters. Existing computational models have mostly focused on simulating ‘circular’ agent clusters. This paper questions the realistic appeal of a circular manifestation for conversational clusters. As a comparative case study, two contrasting models have been implemented: Model 1, like existing approaches, simulates circular agent clusters; whereas Model 2 yields agent clusters of various shapes. Outcomes of the two models are then compared with video data of naturally occurring conversational clusters. Comparison shows that neither Model 1 nor Model 2 are fully self-contained to simulate realistic shapes for agent clusters. Results also demonstrate that circle isn’t the only plausible spatial manifestation for conversational clusters.