Peter Frank Hitchcock
Department of Psychology
Emory University
This talk applies computational models to address two questions core to depression and internalizing disorders broadly: First, how do we judge ourselves negatively? Second, why do we sometimes engage in repetitive negative thinking (RNT), such as rumination and worry? Addressing the first question, in part one I will describe a new task and computational approach to investigate self-judgment. Self-judgment tasks are designed to infer individual differences in latent self-schemas, and elucidate how such schemas are sampled from to form self-judgments. Yet, these tasks tend to confound explicit self-beliefs with the judgment process itself. We drew on methods from value-based decision-making to de-confound these variables. We found that more depressed individuals judge themselves more negatively, beyond what is explained by their explicit self-beliefs. Addressing the second question, in part two I will offer a normative account of RNT from a computational perspective. I will show that a model that learns to gate emotional content into working memory via trial and error exhibits the hallmark features of RNT. This suggests that some degree of RNT may arise as a byproduct of adaptive—but blind and incremental—learning about what emotional content is useful to hold in mind. Together, these projects show how computational theories and tools can offer insight into complex and vexing clinical phenomena.
View a recording of this session here.