Integrative Model for Interoception and Exteroception: predictive coding, points of modulation, and testable predictions

Abstract: Interoception and exteroception provide continuous feedback about the body and the environment, yet how they are dynamically integrated within a unified predictive coding framework has remained under-specified. This paper develops and empirically validates an integrative predictive coding model that treats interoceptive and exteroceptive inference as parallel hierarchical systems exchanging precision-weighted prediction errors. Within this framework, arbitration between the two streams is governed by relative precision weights (w) and integrated within the anterior insula (AIC) and anterior cingulate cortex (ACC). Computational simulations of the model reproduced biologically plausible dynamics: prediction errors decayed exponentially while arbitration weights self-normalized toward equilibrium (w = 0.5), demonstrating stable convergence and coherent integration. Simulated anxiety and PTSD profiles, characterized respectively by interoceptive and exteroceptive overweighting, yielded rigid, self-sustaining imbalances (w to 1 or w to 0) and slowed recalibration. Empirical application of the arbitration equation to published EEG-fMRI datasets further validated the model. The framework contributes a unifying account of how dysregulated precision weighting may underlie anxiety (overweighted interoception) and PTSD (underweighted interoception). Building on this validation, a proposed experimental paradigm is outlined to test the model's predictions in humans. It examines recalibration across anxiety, neutral, and PTSD groups following targeted interoceptive or exteroceptive therapies. Key predictions include identifiable neural markers of coherence, modulation of heartbeat-evoked potentials by vagal stimulation, and precision-sensitive behavioral signatures in interoceptive-exteroceptive congruency tasks.