Belonging-Gated Learning: A Missing Control Parameter in Predictive Neuroscience

Learning is often modeled as a function of prediction error minimization, attention, or reinforcement. However, these models routinely fail to explain why learning collapses under social threat, exclusion, or loss of psychological safety—even when sensory data and incentives remain intact. This preprint introduces Belonging-Gated Learning, a formal extension of predictive neuroscience and the BEP (Belonging–Energy–Prediction) framework. It argues that Belonging (B) functions as a necessary control parameter for sustained learning and belief updating. When belonging falls below a critical threshold, predictive systems do not merely slow or bias learning—they undergo a phase transition in which learning is actively suppressed and cognition defaults to low-cost coherence modes (e.g., conformity, narrative preservation, or synchrony). The paper integrates findings from: predictive processing and active inference social neuroscience and attachment theory autonomic regulation and energy constraints education, therapy, and human–AI interaction Belonging-Gated Learning reframes learning not as a purely informational process, but as a field-dependent phenomenon constrained by social safety and energetic capacity. This model explains why exclusion, shame, and threat reliably impair learning across domains, and why restoring belonging rapidly reopens epistemic flexibility. The work positions belonging as a measurable, testable variable in learning dynamics, with implications for neuroscience, education design, psychotherapy, organizational behavior, and AI alignment. It serves as a conceptual bridge between classical predictive models and emerging field-theoretic approaches to cognition.