Journal of Psychiatric Insight Review

Abstract

Background: Relapse is the central unsolved problem of addiction medicine, with return-to-use rates commonly estimated at 40 to 60 percent within a year of treatment. The dominant neurobiological accounts, including the brain-disease model, incentive-sensitization theory, the allostasis (“dark-side”) model, the actions-to-habits-to compulsions framework, and the impaired-response-inhibition-and-salience-attribution (iRISA) model, powerfully explain how vulnerability develops and endures, yet each treats relapse largely as the re-expression of a stable trait. They under-specify the acute question that most concerns clinicians and patients: why does a particular person return to use at a particular moment?

Methods: This is a conceptual, integrative synthesis. Drawing on computational neuroscience (active inference and predictive processing), interoception research, addiction neurobiology, and the clinical relapse-prevention literature, we integrate seminal and recent work into a single mechanistic account and derive testable predictions.

Results: We advance the Precision-Collapse Model (PCM), which reframes relapse not as reactivation of disease but as a transient, state-dependent failure of precision control: a momentary collapse in the precision weighting of prediction errors that permits entrenched maladaptive priors (“using will relieve this”) to dominate policy selection. Canonical relapse triggers (stress, negative affect, cue exposure, sleep loss, and interoceptive perturbation) are reinterpreted as sharing one computational signature: each transiently biases the precision balance toward aversive interoceptive prediction errors and away from exteroceptive and goal-directed counter evidence. Incentive-sensitization, allostasis, habit and compulsion, and iRISA are shown to be special cases of this dysregulation.

Conclusion: PCM shifts the unit of analysis from disease to inference and from trait to event, reframes mindfulness based and interoceptive therapies as “precision retraining,” and specifies a roadmap for a computational relapse signature combining interoceptive assessment with ambulatory sensing. It reduces stigma by casting relapse as a predictable dynamical event rather than moral or biological inevitability.