Semantic Debt

This paper proposes Semantic Debt as a standalone ontology of meaning failure in software, governance, and agent-built systems. Semantic debt is defined as the accumulated future cost of meaning that is underdefined, misplaced, contradictory, stale, fragmented, opaque, overcoupled, overclaimed, inadmissible, or only recoverable through inference. Unlike technical debt, which concerns implementation compromise, semantic debt concerns the recoverability and admissibility of meaning itself. The paper's central contribution is a two-axis ontology. The first axis defines seven semantic subdomains: lexical debt, ontological debt, boundary debt, procedural debt, evidentiary debt, authority debt, and claim debt. The second axis defines nine debt modes: underdefinition, contradiction, fragmentation, staleness, opacity, overcoupling, inadmissibility, overclaim, and lossy translation. Together, these axes form a diagnostic matrix for locating semantic debt by both the kind of meaning that is failing and the failure pattern involved. The paper gives special attention to semantic fragmentation as a cross-cutting distribution failure, and identifies semantic locality debt as the precise subtype in which meaning exists somewhere in a system but lacks a stable, canonical, locally recoverable home. It then connects the ontology to operational governance artifacts: glossaries, non-claims files, claim ledgers, standing graphs, migration contracts, receipt schemas, and semantic debt ledgers. The framework is intended for software systems, AI-assisted development, governance protocols, long-lived infrastructure, and any system whose claims must remain recoverable, bounded, and admissible across change.