Synthetic Coherence: Truth Fidelity and the Signal Ecology of Representational Systems

Representational systems, human, institutional, and computational, can maintain internal coherence while losing contact with the reality they claim to describe. This working paper introduces synthetic coherence (SC) as a diagnostic term for that condition and proposes a cross-domain architecture for detecting it through diagnostic failure sites. It separates truth fidelity, answerability, and governing coupling as distinct properties, outlines complementary assessment families (empirical constraint-return and structural adequacy analysis), and demonstrates the framework through institutional (Challenger) and AI (LLM benchmark optimisation) case study examples. The paper includes recent evidence on AI-mediated coherence amplification at the levels of belief validation, information discovery, and authorship.