A Coherence-Gated Bayesian Decision Rule

Abstract

Raoul Bianchetti and Payam Danesh

In this paper, we formulate a coherence-gated Bayesian decision rule within Viscous Time Theory (VTT) frame. A standard posterior is multiplied by a bounded gate constructed from alignment, retained coherence cost, informational support, effective informational viscosity, and local sensitivity. The result is normalized only when its total gated mass is positive; otherwise, the rule abstains. We prove boundedness and well-posedness, exact recovery of the Bayesian posterior at the neutral gate, an explicit perturbation bound, an odds decomposition, an exact Bayesian representation through an auxiliary admissibility event, and stability away from zero gated mass. A fully specified three-class synthetic study separates predictive features from an observed support-quality channel. Across convergent, bifurcation, and collapse regimes, coverage decreases from 89.3% to 15.3%, while accuracy on covered cases remains between 98.0% and 99.4%. These values demonstrate the intended selective behavior only under the stated construction.

PDF