Agent-Based Behavioral Modeling of Human Associative Learning in a Complex Approach-Avoidance Conflict Task

Abstract Despite its key role in the development, maintenance, and treatment of anxiety disorders, the detailed mechanisms of human avoidance learning remain elusive. To contribute to the understanding of avoidance learning, we here report on a novel approach-avoidance conflict task that requires participants to learn associations between complex visual stimuli and combined appetitive and aversive stimuli while actively engaging with the experimental environment. Using an agent-based behavioral modeling approach, we implemented and validated an extensive set of control, heuristic, Rescorla-Wagner learning-based, and hybrid agents. We show that a Rescorla-Wagner learning-based agent with a prior expectation bias parameter best explains the learning behavior of 50 participants. As such, our work complements current research on the computational underpinnings of approach-avoidance behavior by showing paradigm and task instruction dependencies in approach-avoidance-relevant associative learning and contributes to the overall aim of achieving a more fine-grained understanding of the etiology of anxiety disorders.