Energetic and Regulatory Constraints on Predictive Inference: A Thermodynamic Extension of Computational Psychiatry

Predictive coding models have provided a robust algorithmic description of psychopathology as a failure in hierarchical Bayesian inference. However, these frameworks often lack biophysical closure regarding the energetic costs associated with information processing. This paper proposes a minimal thermodynamic extension to computational psychiatry by introducing two system-level parameters: Entropic Overhead (phi) and Dissipative Capacity (alpha). We argue that the high-dimensional state space of the human neocortex imposes metabolic demands that act as fundamental constraints on neural stability. By formalizing a stability condition where the rate of uncertainty reduction must not exceed the system's dissipative limit, we identify boundary conditions under which predictive inference becomes inherently unstable. We discuss how recent empirical observations—such as connectivity-dependent neuromodulation efficacy and resting-state gamma dysregulation—are structurally consistent with this thermodynamic regulatory model. Integrating these constraints allows for a more complete mechanistic account of psychiatric vulnerability as an emergent property of high-dimensional neural systems operating under metabolic efficiency constraints.Python implementation of the Neurodynamic Rigidity (NR) Index:https://github.com/Dasein193/Neurodynamic-Rigidity-Index