Resolution-Based Information Theory: Degradation Control as a Design Principle for Multi-Agent System Stability

Abstract

Resolution-Based Information Theory examines how declared generative constraints can be preserved when receivers have limited and changing capacity. It distinguishes task-specific discrimination, intent preservation, and local recoverability. Version 2.0 treats the signed sender-receiver accuracy gap as a candidate mismatch indicator whose link to overload requires independent testing. Conditional drift arguments bound exit or regime-ending correction times, while nested sets distinguish loss of recovery reserve, intent preservation, and returnability. These mathematical conditions do not follow from the accuracy gap alone. Historical predictions and prospective instruments are separated; the measurement bridges, combined detector, and operational upscaling certificate remain unvalidated.