Errors or Adaptations? A Critical Review of Predictive Processing in Psychiatry

Predictive processing (PP) accounts often characterize mental illness as maladaptive and epistemically distorting due to mismatches between brain-generated top-down models and bottom-up sensory inputs, with this review identifying exceptions. First, hypervigilance in trauma survivors with PTSD or depression may sustain desirable gaps between anticipated problems and actual harms that would otherwise occur. Second, PP defenders have argued that depressive slowdowns follow from maladaptive brain-based regulatory models, yet physiological problems may make activity strenuous—in which case slowing down is adaptive. Third, PP researchers introduce tacit normative assumptions. For example, in autism and ADHD, they stipulate thresholds for how specific (hence error-prone) predictive models should be, and PP interpretations of schizophrenia sometimes presuppose Western concepts of self as normative neurocognitive ideals. Fourth, PP accounts of prediction error can tacitly invoke veridical representation, even though advocates regularly claim that cognition evolved primarily for action, not truth-seeking. While criticizing PP for its overreaches, this review also explores how greater attention to these exceptions and factors such as cultural variability may strengthen the framework’s capacity to understand and contribute to the treatment of a range of psychiatric conditions.