Conference on Cognitive Computational Neuroscience

Flexibility is a defining property of model-based planning—agents who plan can quickly respond when tasks or circumstances change. However, classic findings in psychology demonstrate that even when fully engaged in planning, humans can sometimes display surprising rigidity in how they conceptualize problems (e.g., Einstellung effects, or “functional fixedness”) that can produce impasses in problem solving. Here, we develop a computational account that may help explain this puzzling combination of flexibility and rigidity in terms of construal sets. Specifically, we present a formal account of hierarchically selecting a “construal” (an ad hoc planning model), in which a decision-maker first selects a construal set (e.g., whether to think of a task in coarse-grained or fine-grained way) and then engages in explicit reasoning to select a specific construal from that set for a task. Simulations show how rigidity in the type of construal used can emerge. In ongoing work, we are testing these predictions with human participants.