Attention in Virtual Environments: Behavior in Locations Shapes Spatial Connectivity

Unlike connections of physical spaces typically observed through movement trajectory, connections of virtual spaces can be observed in high resolution through vision-based trajectory. Studies have explored user movements and their diffusion in virtual environments (VEs). However, frameworks are needed to objectively measure vision-based behaviors in VEs and characterize spatial networks. This paper proposes a spatial mapping approach for user data through visual-spatial attention, using cognitive psychology and spatial ecology concepts. The aim was to demonstrate a computational tool that can mathematically describe the connectivity of locations based on the trajectory of user attention. We first implemented VEs to represent specific places, which were then sampled into cells, collected user location and direction data within the VEs and calculated attention areas using the direction vector, constructed spatial graphs by creating links among cells within the attention area, and finally calculated the centrality of nodes within each spatial graph and performed community detection. The tool was tested on log data from two user studies on VEs. In the results, the centrality indicated the cells where user attention was focused, and the community detection identified cell clusters. By analyzing the features of cells with high centralities, such as buildings, lakes, and non-player characters, we can identify cells with similar features that will attract attention in VEs. Using the proposed tool, a quantitative description of attention was obtained without direct feedback from users. Practically, the spatial graph can provide guidelines for designing specific areas to attract attention and help managers cluster cells for management.