The goal of this proposal is to characterize novel computational biases in learning and decision-making in PTSD. The dominant understanding of PTSD emphasizes heightened fear learning and threat detection, and weakened fear extinction / inhibition. While this model explains many aspects of PTSD, there is a growing gap between our models of PTSD and the emerging literature defining the normative computational mechanisms of learning and decision-making. This proposal aims to bridge this gap by defining computational biases in two clinically-relevant domains of learning and decision-making in PTSD. First, the mechanisms by which individuals with PTSD prefer avoiding threat at the expense of losing potential reward is not understood. This bias in approach-avoidance conflict resolution is an essential feature of the clinical presentation of PTSD, and though threat processing and reward processing have separately been characterized in PTSD, how threat and reward processing interact to result in biases towards avoidance has never been investigated. Second, dysregulation of context-depending (i.e., latent state) learning has clear clinical implications in PTSD: generalization of threat learning outside the trauma context is related to the development of PTSD; generalization of extinction learning outside of the clinical context is related to the treatment of PTSD. However, computational models of context-modulated learning have not been used to understand these processes in PTSD. The current project proposes to use computational modeling of learning and decision- making in novel tasks that probe the behavioral and brain mechanisms of approach-avoidance biases (Specific Aim 1) and context-modulated (i.e., latent state) learning (Specific Aim 2). A case-controlled design would be used, in which healthy adults, trauma-exposed adults without PTSD, trauma-exposed adults with PTSD, and adults with non-PTSD anxiety disorders would undergo novel learning and decision-making tasks during fMRI with concurrent psychophysiological assessment. By defining novel computational biases in learning and decision-making in PTSD, the project 1) would bridge the gap between our understanding of PTSD and our the growing science of computational mechanisms of learning, 2) has the potential to explain clinically-relevant features of dysfunction in PTSD, and 3) would provide targets for tracking trajectories of PTSD development and treatment, and stimulate novel methods for treating PTSD that go beyond the traditional fear conditioning and extinction models.
The proposed project would provide important knowledge regarding computational biases underlying threat-and reward-learning, approach-avoidance conflict, and context-modulated learning and recall in PTSD. This important knowledge would bridge the gap between our understanding of PTSD and the growing science of computational mechanisms of learning. This knowledge also has the potential to explain clinically-relevant features of dysfunction in PTSD, would provide targets for tracking trajectories of PTSD development and treatment, and stimulate novel methods for treating PTSD that go beyond the traditional fear conditioning and extinction models.
