Cognitive Units: A Core Representational Artifact Model for Governed Cognitive Transformations in Operational Systems

Operational systems may retain observations, documents, model outputs, graph relations, provenance records, workflow histories, and human assessments while still losing the representational continuity needed to explain how a recognized state acquired its standing. The problem is therefore not necessarily missing information, but the absence of a stable artifact boundary through which evidence, context, interpretation, authority, temporal applicability, revision, and future-use conditions remain connected. This paper asks how cognitively significant elements of governed transformations and their relationships to evolving recognized states can be represented explicitly in operational systems. It proposes Cognitive Units (CUs) as a core representational artifact model for this purpose. A Cognitive Unit is a persistently identifiable, bounded, and governable artifact that represents a cognitively significant element of a governed transformation and preserves the conditions under which that element can be interpreted, related to recognized state, revised, and considered for future use. The proposed contract combines three artifact-level preconditions with six constitutive responsibilities: cognitive payload, semantic context, governance, traceability, temporal scope, and reuse policy. Five proposed foundational roles—SignalUnit, ContextUnit, StateUnit, MeaningUnit, and MemoryUnit—provide an open starting grammar rather than a processing pipeline or closed taxonomy. The System Corpus supplies the bounded and versionable validity frame in which CUs are interpreted and evaluated, while operational events, cognitive operations, recognized-state transitions, cognitive transformations, and CUs remain distinct representational levels with non-one-to-one relationships. The model provides an artifact-level basis for state and cognitive traceability, reconstructable explanation, governed revision, cognitive continuity, controlled future reuse, inspectable human contribution, heterogeneous composition, and cross-corpus mediation. Two hypothetical transformations provide representational illustration, with complete cases in Appendix A and a non-normative executable realization in Appendix B. The contribution is conceptual and representational; formal profiles, independent implementations, implementation-cost analysis, and empirical evaluation remain future work.