Computational approach to depression: From the viewpoint of reinforcement learning

Abstract

Depression is a highly recognized mental disorder that has been widely studied. However, an essential understanding of depression has not been achieved. We conducted a narrative review to examine the mechanism of depression, based on the perspective of a computational approach. We focused on studies that used the reinforcement learning model. We reviewed the relationship between the parameters of the reinforcement learning model and anhedonia symptoms. The computational approach to depression is a new research field; therefore, we will also propose future topics for study on the basis of our narrative reviews. In particular, we discuss increasing the quality of our findings by using model-based reinforcement learning and the Bayesian inference models, and exploring the mechanisms of psychotherapy.