In Search of a Computational Model of Depersonalization Using a Bayesian Recurrent Neural Network Approach

Mario Zarco

Keywords: Bayesian Recurrent Neural Network, Depersonalization, Computational Modeling, Active Inference, Homeostasis

TL;DR: We train a Bayesian recurrent neural network that integrates multimodal signals to investigate how altered predictive dynamics may contribute to depersonalization.

Abstract: Researchers have speculated that depersonalization is triggered in part by alterations in predictive models resulting from abnormal interoceptive dynamics. However, computational models exploring these mechanisms remain scarce. In this work, we trained a Bayesian Recurrent Neural Network that integrates multimodal signals in a threat-avoidant task, providing a platform to investigate potential pathways underlying this condition.

Submission Number: 6