Computational modeling is gaining increasing popularity in the field of psychiatry. By applying a machine learning framework known as reinforcement learning (RL), researchers in the domain of computational psychiatry have begun to shed light on the latent decision-making processes that are altered in mental disorders. In this chapter, we first explain how RL can be used to characterize the neurocomputational processes of decision-making and learning. To illustrate this approach, we focus on a set of studies that employ a basic class of RL, known as model-free RL, to gain insights into the psychopathology of obsessive–compulsive and gambling disorders. Lastly, we review a common model-fitting procedure used in computational psychiatry research and offer a practical guide, also considering potential pitfalls. This encompasses discussions on estimating model parameters (parameter estimation), comparing multiple models based on their goodness of fits (model comparison), and testing the robustness of model-fitting procedures using synthesized data (model and parameter recovery analyses). Given the rapid growth of this area of study, we believe that these discussions will be of great value to authors, reviewers, and readers of computational psychiatry literature.