Theory-Driven Computational Psychiatry and Its Challenges: An Optimistic Outlook

Mental health disorders affect millions worldwide, yet our understanding of their underlying mechanisms remains limited, impeding the development of precise and effective treatments. Computational psychiatry, a burgeoning field at the intersection of neuroscience, psychology, and computational modelling, seeks to bridge this gap. By employing theory-driven computational models to characterise the neural and cognitive processes underlying psychiatric conditions, it provides a quantitative framework that could help reshape how we conceptualise, diagnose, and treat these disorders. Despite its potential, computational psychiatry has received criticism for not meeting these clinical goals due to challenges relating to issues such as model selection and recoverability, posing questions about the robustness and interpretability of the field's insights. Here, we show how each of these challenges can be addressed. We argue that the challenges, and our approaches to addressing them, echo long-standing, general debates in philosophy of science, which are inherent to modelling of complex systems, rather than specific to the use of computational methods in psychiatry. This review provides an optimistic outlook for computational psychiatry, where the field can advance by clearly articulating and addressing the challenges faced.