Main Article Content
Depression is a leading source of medical disability and is experienced by over 322 million people worldwide. Despite its increasingly significant burden and a pressing need for effective treatment, depression has been persistently difficult to treat. Current common practice for treatment selection is an educated guess-and-check approach, in which clinicians prescribe one of the numerous approved therapies for depression in a stepwise manner. Though evidence-based clinical guidelines for managing depression exist, there is a paucity of evidence to support specific treatment recommendations. A significant barrier to developing such recommendations is the symptom heterogeneity present in the diagnosis of major depressive disorder. Machine learning offers the ability to recognize this heterogeneity and model that information in psychiatric disorders. Specifically, machine learning allows processing of high-dimensional data, the management of missing data, creating high- level abstractions, and the freedom of not requiring a priori patient stratification. While ethical concerns arise in employing these methods, the benefits are wide-reaching, from personalizing treatment for depression, to the development of artificial intelligence chats that employ psychotherapy, to predicting social outcomes for patients with mental illness. The implications extend far beyond depression treatment, as the epidemiology and service demand for mental healthcare systems continue to grow. Indeed, psychiatry is primed for innovation in artificial intelligence and machine learning.