Information availability shapes foraging strategy: A virtual foraging game

Understanding spatial exploration reveals how organisms sample, process, and act on spatial information. Foraging paradigms, in which participants search for multiple items in an environment, offer a window into these exploration strategies. In this study, participants foraged as first-person avatars in a virtual environment. We investigated how factors such as movement ability and information availability influence foraging strategy and developed a generative model which predicts sequences of item selections using the latent parameters 'proximity' (selecting items closer to the previous selection), 'momentum' (selecting items in front or doubling-back) and 'turning' (left or right turning tendency). We demonstrate how we can test the importance of these parameters in predicting behaviour, showing that proximity and momentum are highly predictive of foraging behaviour in this task. We also find that that information availability impacts behaviour: participants with a map are more likely to double-back. Finally, we explore how different target selection rules affect model performance, showing that a `winner-takes-all' rule outperforms a stochastic rule and a simple baseline model. We conclude that our model can accurately predict behaviour on an item-by-item basis and argue that our approach can act as a template for understanding the cognitive underpinnings of human visual exploration behaviour.