Understanding dual process cognition via the minimum description length principle

Ted Moskovitz, Kevin J. Miller, Maneesh Sahani, Matthew M. Botvinick

Abstract: Author summary Since the founding of psychology as a scientific discipline, scientists have recognized that much of human and animal behavior falls into two separate categories: slow, effortful deliberation (e.g., solving math problems) and fast, automatic habitual action selection (e.g., signing your name). Such “dual process” theories of cognition have since been developed to explain a number of phenomena from across psychology and neuroscience. Despite all of these experimental results, understanding why this pattern in behavior exists has proven elusive. In this work, we develop a model of dual process cognition which can be derived independently from experimental observations about the brain. Our theory proposes that the division between effortful and automatic behavior emerges as a natural result of the need to balance mastery of the tasks we have to complete every day and flexibility to adapt to new circumstances. In a large number of computer simulations, our model captures many behavioral phenomena associated with “dual process” cognition across multiple subfields of cognitive science.