CRCNS: Computational Foundations for Externalizing/Internalizing Psychopathology

A core question in mental health is what underlying processes give rise to symptoms (e.g., NIMH strategic plan goal #1). Answers may lie less in individual diagnostic categories, but instead in global, transdiagnostic patterns of symptoms, notably their prominent and clinically useful division into internalizing (e.g., anxiety) vs. externalizing (e.g., aggression) forms. This project aims to characterize these two symptom clusters and their development (per NIMH strategic plan goal #2), by relating them to computational mechanisms for decision making that have been studied in the healthy brain. Previous computational psychiatry research grounds some internalizing symptoms such as worry in dysregulated mental simulation, or "internal information seeking." Here, we propose and test a hypothesis to extend this framework to comprise externalizing symptoms, which we suggest are grounded in parallel dysregulation of external information seeking (exploration of the environment), building on a recent theory. Because these computational capacities, as well as many mental health symptoms, emerge in childhood and adolescence, there is a unique opportunity to understand their relationship via development. We will use computational modelling to derive signatures of both sorts of information seeking from participants' choice behavior in two reinforcement learning tasks. We hypothesize that internalizing vs. externalizing symptoms are associated, respectively, with enhanced internal vs. external information seeking, and further reflect aberrant developmental trajectories. We test this in Aim 1 by comparing task behavior to psychiatric symptoms in two large general population samples collected online in adults. Next, in Aim 2, we examine how these processes develop using the same tasks in children and adolescents, and how this development differs in children with a diagnosed internalizing or externalizing disorder. The present research leverages and tests a unifying computational theory that situates both types of information seeking as parallel options in a tradeoff between acting for immediate reward vs delaying to gather information and improve later choices. This account can overcome a crucial gap in current computational psychiatry research, which only accounts for a relatively narrow range of symptoms. By connecting computational neuroscience, psychiatry, and development, this project will clarify the neurocomputational foundations of a wide range of externalizing and internalizing symptoms.