Computational Modeling of Interoceptive Perception across the Psychosis-Spectrum

Project Summary/Abstract Distressing psychosis symptoms (i.e., hallucinations and delusions), exist on a spectrum from subclinical to those observed in diagnosed individuals, occur in an astonishing 5-8% of the general population, and are often accompanied by disability, reduced social functioning, and cognitive impairments. Mechanisms underlying psychosis-spectrum symptoms remain unclear, limiting development of targeted treatments. A new approach focuses on internal body states (e.g., heartbeats, temperature) that shape experiences of self, others, and the world, and are essential to adaptive behavior. Interoception encompasses the detection, integration, and interpretation of internal body states, which appears disrupted and critically understudied in psychosis. Clinical exemplars include mismatched behavior (e.g., wearing a parka in hot weather), reality distortions (e.g., believing back pain is from a tracking device), and loss of body ownership (e.g., perceiving external control). Interoceptive accuracy tasks reveal poor detection of internal body states across the psychosis-spectrum; however, this cannot be separated from cognition. To move beyond descriptive work and limitations, the current study non-invasively manipulates internal body states in a predictive coding paradigm. Further, computational modeling dissects interoceptive predictive coding into subprocesses and disentangling relationships with psychosis symptoms versus cognition to provide translational insights that may inform future treatment targets. Predictive coding is a neuro-cognitive framework, where ‘prior beliefs’ allow for predictions about incoming sensory information for efficient responding. A prediction error results from a discrepancy between the ‘prior belief’ and sensory information, which should update the belief for future adaptive behavior. Therefore, the brain makes sense of the world through iterative hypothesis testing, by comparing ‘prior beliefs’ (top-down predictions) with sensory information (bottom-up stimuli). This framework has provided valuable insight to exteroceptive perception (e.g., auditory, visual) in psychosis-spectrum samples, where strong ‘prior beliefs’ dominate perception. Similarly, strong interoceptive ‘priors’ may drive poor interoceptive accuracy in psychosis, and be a critically untested process contributing to both false perceptions (hallucinations) and formation of false beliefs (delusions) that fail to reflect the environment. Computational models of predictive coding can dissect interoceptive perception into higher- and lower-level subprocesses, including weighting of ‘prior beliefs’, adaptability, and behavioral consistency, to test differential relationships with key clinical features of psychosis. Thus, the current study will move the field forward by 1) testing a hypothesis of strong interoceptive ‘prior beliefs’ within the psychosis-spectrum, and 2) investigating differential relationships for subprocesses of interoceptive predictive coding with psychosis symptoms versus cognition. Together, the proposed research and training plan establish a critical foundation integrating cognitive science theory and computational methods with basic interoception research to advance our understanding and treatment targets for distressing psychosis symptoms.