Observer-Conditioned Semantic Transport

Meaning is often treated either as an intrinsic property of a signal or as a task-specific quantity defined independently of the receiver. This paper develops an observer-conditioned alternative. An observer o in context c is represented by an interpretation kernel K_o(dm|x,c) that maps a signal x to a distribution over a declared meaning or task-relevant interpretation variable m. Two signals are semantically equivalent for that observer when they induce the same interpretation law; different observers can therefore impose different semantic partitions on the same signal space. A semantic bridge is a transformation chosen to minimize decision-relevant or distributional distortion between observers. “Semantic resonance” is operationalized as unusually efficient transfer under a fixed communication, sample, or adaptation budget rather than assumed physical frequency matching. “Semantic osmosis” is reformulated as convergence of observer-conditioned interpretation structures under repeated coupling, with internal detectability treated as a separate observation problem rather than built into the definition. The framework connects observer-relative identifiability, semantic information, information bottleneck ideas, and semantic communication while remaining distinct from a 2026 lexical Semantic Field Theory based on contextual deformation and energy stabilization. We give exact equivalence and bridge-composition results, identify the assumptions needed for metrics and dynamics, and propose synthetic-world experiments with language models that can falsify the central claims.