Model-based planning in structured foraging environments

In order to maximize reward, humans need to balance engaging with currently available sources of reward and searching for better ones. Optimal foraging theory provides a formal but simple mathematical choice rule to make such stay/leave decisions. However, this rule (given the Marginal Value Theorem; MVT) assumes that such decisions rely on a strategy that does not consider the structure of the environment. In other words, it does not leave room for planning during foraging. Yet, the real world is replete with such opportunities. Therefore, we developed a new structured foraging task to study how people employ goal-directed planning during foraging. Specifically, we explore the extent to which participants incorporate an internal model of the task structure during stay/leave decisions. We find that this task elicits behavior that deviates from the optimal decision rule in standard foraging tasks (MVT): participants use the task’s structure to consider the value of alternative reward options when deciding to leave their current one. Importantly, this behavior is pronounced in more goal-directed participants. Computational modeling suggests that these deviations are beneficial, but to an extent dictated by choice stochasticity. This study provides a novel method for studying decision making in structured environments, and has implications for understanding how foraging and planning interact.