Unraveling neural mechanisms underlying learning in a noisy, dynamically changing world

About This Grant

Project Summary Uncertainty-related cognitive dysfunctions are central to anxiety disorders, behavioral addictions, attention- deficit hyperactivity disorder, and schizophrenia, yet remain poorly understood. Anxious individuals, for example, are highly intolerant of uncertain situations, while those with gambling addictions often seek uncertainty, leading to loss-chasing behaviors. These diverse clinical presentations may stem from a fundamental computational challenge: the brain must simultaneously distinguish between two types of uncertainty—moment-to-moment stochasticity of observations and environmental volatility (how quickly underlying causes change)—that require opposite learning strategies. Previous work has focused on one factor or the other, but in reality, both volatility and stochasticity are unknown and potentially changing. Importantly, while they both increase experienced noise, they require opposite behavioral responses, making their dissociation both critical and computationally difficult and prone to systematic errors. Our recent work provides a computational framework for how the brain solves this challenge and how this process breaks down in psychiatric illness. In a large-scale neuroimaging program spanning three aims, we combine behavior, computational modeling, simultaneous fMRI-pupillometry, and causal arousal manipulation to elucidate neural mechanisms processing uncertainty while systematically manipulating both volatility and stochasticity across different outcome types. We will test specific hypotheses about the neurocomputational mechanisms of uncertainty processing (Aim 1), determine whether they are causally linked to arousal mediated by the locus coeruleus–norepinephrine system (Aim 2), and examine how these processes differ across outcome types (Aim 3), providing a mechanistic foundation for understanding uncertainty-related symptoms across psychiatric disorders.