Abstract
Visual foraging has been used in research as a controlled proxy for spatial exploration in humans. In one common paradigm, participants search for two different classes of targets on a 2D display. Our recent work has developed a state-of-the-art model that can predict the particular sequence of targets a person will select during this task and formalises the latent parameters that underlie the behavioural processes, such that behaviour can be characterised in a cognitively meaningful manner. For example, we have recently extended our model to include absolute direction information (in addition to relative direction) and have shown that there is a strong preference for using cardinal directions when completing 2D screen based tasks. Here, we further extend this model to a 3D team foraging task. In this task, participants controlled a human avatar to navigate a virtual environment and coordinated with their 3-player team to collect as many targets as possible. We extend our framework to work within a 3D multiplayer virtual environment, and show that the influence of spatial factors such as target proximity and momentum (i.e. whether participants continue in a straight line, or double back on themselves) become much more important than in a 2D task. Finally, by modeling multiple participants we investigated how a ‘teaming’ parameter that models how closely different team members track each other varies with task instructions. These results show that our generative model allows us to flexibly model relatively complex exploration behaviour with a small number of psychologically relevant parameters and as little as one trial of data.