Cognition as Information-Theoretic Compression: Non-Convergence, Truth Expansion, and Social Recompression

Human cognition operates under severe informational and energetic constraints. While ob- jects admit arbitrarily fine-grained description, linguistic representation does not converge to a complete account. Nevertheless, finite agents successfully recognize, act, and communicate. This paper proposes that cognition should be understood as an information-theoretic compression process: a selection of minimal representations sufficient for functional adequacy. We argue that linguistic description is non-convergent due to indefinite decomposability of objects across structural, qualitative, and contextual dimensions. In contrast, cognition enforces a stopping condition by minimizing informational cost under task-dependent constraints. At the individual level, this results in intrinsic limits of recognition; at the social level, it yields incompleteness in shared meaning. Truth is then reinterpreted not as a primary target of cognition, but as a high-cost represen- tational expansion that emerges when compression constraints are relaxed, typically under con- ditions of surplus resources. However, such expanded representations cannot function socially without undergoing recompression into transmissible and operationally usable forms. Thus, truth acquires value only through its reintegration into socially shared, compressed structures. By unifying linguistic non-convergence, cognitive limitation, and the social circulation of knowledge within a single framework, this paper offers a structural account of cognition as compression, truth as decompression, and society as recompression.