Interpreting Anxiety Disorders From the Perspective of Interoceptive Computational Models

Interoception-the nervous system's sensing, integration, and interpretation of internal bodily signals-facilitates the dynamic alignment of internal states with external environments. Predictive processing theory posits that this alignment arises from iterative comparisons between predicted and actual sensory inputs. Persistent mismatches in these computations may drive interoceptive dysfunction, a core mechanism implicated in anxiety disorders. Despite advances, a significant gap remains in linking computational models of interoceptive dysregulation to clinical interventions. This review synthesizes evidence to propose an interoceptive computational framework to bridge mechanistic insights with therapeutic innovation for anxiety.

This review synthesizes evidence from computational psychiatry, neurophysiology, and clinical studies to model anxiety as a disorder of interoceptive prediction. We integrate predictive coding mechanisms underlying threat perception with the potential of experimental paradigms and bidirectional modulation strategies for intervention.

Anxiety pathophysiology is driven by hyperprecise threat priors and context rigidity, which amplify interoceptive prediction errors. These computational failures manifest as exaggerated defensive responses, cognitive biases, and maladaptive behaviors. Integrating computational modeling with targeted interventions, such as interoceptive exposure grounded in Bayesian belief updating, improves diagnostic precision and therapeutic outcomes.

By bridging computational theories of interoceptive dysregulation with clinical practice, this framework advances a multidimensional approach to anxiety disorders. Future research should prioritize perturbed-prior experiments and hybrid interventions to optimize personalized treatment. Such integration holds transformative potential for precision psychiatry, addressing both neural computations and embodied experiences in anxiety.