The W3C OWL 2 Recommendation (2012) provides formal semantics based on the description logic SROIQ(D), with three profiles trading expressiveness for tractability: OWL 2 EL (polynomial classification for large ontologies — basis for SNOMED CT's extensive clinical terminology), OWL 2 QL (database-speed conjunctive query answering through first-order rewriting — basis for large-instance queries), and OWL 2 RL (rule-based forward chaining for business logic integration). Key governance-relevant constructs: property chain axioms for authority delegation (R₁ ∘ R₂ ⊑ R₃), HasKey for decision identity, qualified cardinality restrictions for structural constraints ("every decision requires at least one authority source"), and dual semantics (Direct Semantics for decidable reasoning + RDF-Based Semantics for maximal RDF compatibility). The profile system provides the formal foundation for layered governance architectures matching computational guarantees to operational requirements.

RDF's triple model (subject-predicate-object) is inherently more flexible than relational tables for representing complex governance relationships. Named Graphs — collections of triples identified by an IRI — enable grouping by source, tracking provenance at graph level, managing trust by graph association, and versioning by creating new graphs while preserving originals. JSON-LD serialization enables semantic governance data consumption through standard JSON APIs without requiring full RDF stack adoption — a critical adoption consideration for enterprise governance tools.

PROV-O (W3C, 2013) captures "who did what, when, why, and how" through six core properties (wasGeneratedBy, wasDerivedFrom, wasAttributedTo, wasAssociatedWith, actedOnBehalfOf, used) plus Bundles for recursive provenance. PROV-CONSTRAINTS provides temporal ordering and lifecycle validation. The actedOnBehalfOf property directly models authority delegation chains. But PROV-O provides tracking without governing — it answers "where did this come from?" but not "was this properly authorized, adequately evidenced, and consistent with constraints?" The gap between tracking provenance and governing through provenance is the domain-specific contribution space.

SKOS (W3C, 2009) represents knowledge organization systems — thesauri, taxonomies, controlled vocabularies — as linked data. Concept-centric design (skos:Concept, skos:ConceptScheme, skos:Collection), hierarchical and associative relations (broader/narrower/related), cross-scheme mapping (exactMatch/closeMatch/broadMatch), multilingual labeling (prefLabel/altLabel per language), and documentation properties (definition, scopeNote, historyNote). Design principles: minimal ontological commitment (concepts are not classes), no strict transitivity (allowing real-world vocabulary messiness), open world assumption. Isaac et al. (2013) documented key design choices in Journal of Web Semantics. SKOS provides the formal infrastructure for regulatory vocabulary management — authority registries, compliance term mapping, cross-jurisdictional terminology alignment.

Gruber (1993; see also Gruber 1995) in the ontology-engineering literature provided the foundational definition and five design criteria: clarity (objective, documented definitions), coherence (internal consistency), extendibility (anticipated + unanticipated uses), minimal encoding bias (knowledge-level specification independent of implementation), and minimal ontological commitment (fewest claims sufficient to support intended knowledge sharing). The crucial concept of "ontological commitment": an agreement to use vocabulary consistently with a theory specified by the ontology. Ontologies are social contracts about terminology — adopting an ontology means committing to shared conceptualization, not asserting that the world matches the model. This pragmatic framing transforms governance protocol adoption from software installation into ontological commitment — agreement to represent governance concepts using shared vocabulary and axioms.

Davis, Shrobe & Szolovits (1993) in AI Magazine argued that KR serves five roles: (1) Surrogate — the representation stands for the real thing (a decision record is always an imperfect representation of an actual decision; what is omitted is a governance choice), (2) Set of ontological commitments — the choice of primitives determines what the system can "see" (if authority is a primitive, the system can reason about authorization), (3) Theory of intelligent reasoning — the representation defines sanctioned inferences (what counts as valid governance reasoning), (4) Medium for efficient computation — the representation must support tractable queries (governance queries must return in bounded time), (5) Medium for human expression — the representation must be comprehensible to human users (accountability requires that governance conclusions be human-interpretable). This framework reveals that governance representation choices are not merely technical but constitute a theory of governance — the ontology doesn't just store governance information, it embodies what governance means.

BFO (Arp, Smith & Spear 2015, ISO/IEC 21838-2:2021) categorizes entities into Continuants (persist through time) and Occurrents (unfold through time), with further divisions into independent/dependent continuants, processes, and temporal regions. BFO anchors hundreds of biomedical ontologies through the OBO Foundry, achieved ISO standardization, and provides the integration framework for the Industrial Ontology Foundry (manufacturing) and Common Core Ontology (general purpose). DOLCE (Masolo et al. 2003) takes a descriptive/cognitive orientation — modeling how humans conceptualize reality — with the Descriptions & Situations extension enabling formal modeling of contextual interpretation. SUMO (Niles & Pease 2001) is among the largest formal upper ontologies, with deontic categories (Obligation, Permission, Prohibition) directly relevant to governance. GFO (Herre 2010) distinguishes material, mental, and social levels of reality — treating governance structures as real social objects, not mere linguistic conventions.

The gap: BFO has been successfully instantiated for biomedicine, manufacturing, and general concepts. No comparable instantiation exists for organizational decision governance. The domain is modeled procedurally (workflow systems, BPM) but never ontologically (formal knowledge representation with decidable reasoning).

Guarino & Welty (2002) in Communications of the ACM introduced four meta-properties for evaluating ontological choices: rigidity (+R: every instance necessarily has the property), identity (+I: the property provides individuation criteria), unity (+U: a common unifying relation binds parts), and dependence (+D: instances require other individuals' existence). Validation rules: anti-rigid properties cannot subsume rigid ones; identity carries to all subclasses; non-unity cannot subsume unity. OntoClean provides formal quality assurance for governance ontology design — every class in the hierarchy can be validated against meta-property constraints, preventing taxonomic errors like making an anti-rigid class ("active policy") subsume a rigid class ("legal entity").

The DL family ranges from ALC (EXPTIME-complete) through SHIQ and SHOIQ to SROIQ(D) underlying OWL 2 DL (2-NEXPTIME) (Baader et al. 2003; W3C OWL 2, 2012). DL-Lite (basis for OWL 2 QL) achieves AC₀ conjunctive query answering — equivalent to SQL evaluation — enabling database-speed governance queries over large decision datasets. EL++ (basis for OWL 2 EL) achieves polynomial classification. This research establishes that governance system designers must explicitly choose what governance questions the system can answer efficiently: authority chain queries (graph traversal — tractable), compliance checking (instance classification — polynomial in EL++), and complex constraint satisfaction (full DL reasoning — potentially expensive). The choice of description logic is a governance design decision.

Allen (1983) in Communications of the ACM provided 13 mutually exclusive temporal relations between intervals (before, after, meets, overlaps, starts, during, finishes, and their inverses plus equals), with a composition table enabling temporal constraint propagation. Combined with W3C OWL-Time (2017) providing formal temporal concepts (Instant, Interval, Duration), this enables governance-critical temporal reasoning: Was this decision made during the authority's validity period? Did evidence gathering precede the decision? Does constraint applicability overlap the decision context?

Noy & McGuinness (2001) provided a seven-step practical methodology with competency questions as scoping mechanism. METHONTOLOGY (Fernández-López et al. 1997) provides a structured lifecycle: specification, conceptualization, formalization, integration, implementation, maintenance. NeOn (Suárez-Figueroa et al. 2012) addresses networked ontology engineering with nine scenarios for construction through reuse and alignment. Gangemi (2005) established Ontology Design Patterns — reusable solutions for recurring modeling problems including the Situation pattern (contexts as first-class entities) and Description & Situation (contextual interpretation). The broader ODP community contributed the N-ary Relation pattern (W3C SWBPD, 2006) for multi-participant relationships and Provenance patterns for origin/history tracking. These patterns directly apply to governance: decision-making is inherently n-ary (relating decision-maker, authority, evidence, constraints, context, outcome), governance contexts require the Situation pattern, and regulatory interpretation requires D&S.

Berners-Lee (2006) articulated four principles: use URIs as names, use HTTP URIs for lookup, provide information via standards, include links for discovery. The 5-star deployment scheme (★: available → ★★★★★: linked to others' data). Euzenat & Shvaiko (2013) provided comprehensive ontology matching framework: equivalence, subsumption, disjointness, and overlap alignments using string-based, structure-based, instance-based, semantic-based, and background-knowledge techniques. Cross-organizational governance requires both: Linked Data for making governance records discoverable and navigable, and ontology alignment for establishing that governance concepts across organizations have equivalent meaning — not just equivalent data structures but equivalent normative force.

Enterprise knowledge graphs implement a three-layer architecture: storage (graph databases), semantic layer (ontologies + reasoning), and consumption (applications + AI). Graph RAG (2024-2025) extends this by grounding LLM retrieval in knowledge graph structure — reducing hallucination through structural constraints, providing explainable retrieval paths, enabling multi-hop reasoning, and ensuring consistent entity references. The OMG Enterprise Knowledge Graph Task Force (established 2023) signals mainstream enterprise recognition. The application gap: enterprise knowledge graphs are passive — they represent what is, not what should be. They answer "what happened?" not "was what happened authorized?" The transition from passive knowledge graph to active governance substrate — from ontology as documentation to ontology as control plane — is the novel application.