Epistemic overclosure is the condition in which a system’s predictive model ceases to track external dynamics while maintaining internal coherence.
It is the characteristic consequence of informational malfunction. Where malfunction describes the process — asymmetric optimization of internal over external coherence — overclosure describes the resulting state. The system has closed around its own belief structure to the point where new information from the environment cannot penetrate.
The temporal signature is diagnostic: internal KL divergence approaches zero (∂t D_KL(b{t+δ} ‖ b_t) ≈ 0) while external divergence grows (d/dt D_KL(E_t ‖ b_t) > 0). The system is stable by its own lights but increasingly surprised by reality.
Overclosure is not the same as ignorance. An ignorant system has low mutual information with its environment and knows it (or at least has no stake in denying it). An overclosed system has high internal mutual information that actively crowds out environmental signal. The internal model is rich, detailed, and wrong — and its richness is what prevents correction, because every incoming datum is reinterpreted through the overclosed model before it can challenge it.
The term borrows from the logical concept of deductive closure but applies it critically: where deductive closure is a desirable formal property, epistemic overclosure is a pathology in which closure has exceeded its proper scope.