Planning Models for Two-Way Avoidance and Reversal Learning

Reinforcement learning models can explain various aspects of two-way avoidance learning but do not provide a rationale for the relationship found between dynamics of initial learning and dynamics of reversal learning. Artificial intelligence planning, however, offers a novel way to conceptualize the animal's cognitive processes by providing an explicit representation of and reasoning about internal processing stages. Our hybrid planning and plan repair approach demonstrates that the empirically found relationships could be motivated from a consistent theoretical framework.