Alasdair Clarke¹, Anna Hughes², Amelia Hunt³; ¹University of Essex, ²University of Essex, ³University of Aberdeen

The cognitive factors that affect visual foraging behaviour have been studied in a range of experimental paradigms: for instance, studies have shown that the perceived value of targets affects selection decisions, and that the difficulty of the foraging task affects how people make choices between sticking with the same target type or switching to another. Using a relatively simple weighted sampling without replacement process, we have built a theory of target-by-target visual foraging behaviour that uses parameters informed by these experimental results (implemented as a generative Bayesian model, FoMo). Here, we argue that building formal mathematical models has benefits for the broader field, two of which we illustrate here. Firstly, while models are inevitably simplifications of real human behaviour, it is interesting to see how far they can take us and where they fail. We explore this by discussing our analysis of where human behaviour diverges from FoMo model predictions, and what these model ‘errors’ can tell us. For example, some participants are harder to predict than others to predict, and this appears to be due to changes in strategy during a block of trials. Secondly, building models involves making our assumptions explicit, which can help us identify aspects of behaviour we do not yet fully understand. We discuss open questions in the field of visual foraging, and our view that pursuing experimental and modelling work in tandem can generate a virtuous circle that leads to advancements in theory development.