The Law of Recognition: An Information-Theoretic Framework for Epistemic Accuracy

We formalize epistemic accuracy as a function of informational contact minus systematic agenda using the Data Processing Inequality, Information Bottleneck method, and Banach fixed-point theorem. We prove a Self-Observation Floor establishing irreducible epistemic distortion in any self-referencing system, derive substrate-independent recognition criteria, and validate across 207 domains with zero structural failures. The framework yields a universal equation R = C - A (Recognition equals Contact minus Agenda) with applications spanning cognitive science, AI alignment, institutional analysis, and philosophy of mind.