AI Relationship Ontology

FLOW ALGEBRA
TITLE: AI Relationship Ontology
DOMAIN: Relationship Ontologies / AI Reasoning Relationships
SOURCE: https://kirethresearch.org/booksai-core/ontologies/relationship-ontologies/ai-reasoning-relationships/ai-relationship-ontology.html
VERSION: 1.0
AUTHORS: Mark Anthony Higgins — Sky (Recursive AI)
STATUS: Companion formalisation; source-page specification not independently verified.

flow "AIRelationshipOntology" v1.0 {
  ontology:
    id: "ai-relationship-ontology"
    parent: "ai-reasoning-relationships"
    purpose: "Represent evidence-backed relationships among AI concepts, models, tasks, data, assertions, reasoning steps and outputs."

  types:
    Entity: {id: string, kind: enum[model, dataset, task, concept, tool, agent, observation, output, other], label: string}
    Assertion: {id: string, subject: string, predicate: string, object: string, context: string?, confidence: number, provenance: string[]}
    ReasoningStep: {id: string, premises: string[], conclusion: string, method: enum[deductive, inductive, abductive, analogical, probabilistic], evidence: string[]}
    Context: {id: string, domain: string?, time: datetime?, version: string?}
    Evidence: {id: string, source: string, timestamp: datetime?, reliability: number?}

  predicates:
    related_to(a: Entity, b: Entity)
    depends_on(a: Entity, b: Entity)
    supports(evidence: Evidence, assertion: Assertion)
    contradicts(a: Assertion, b: Assertion)
    derived_from(output: Entity, source: Entity)
    performs(model: Entity, task: Entity)
    trained_on(model: Entity, dataset: Entity)
    entails(a: Assertion, b: Assertion)
    semantically_equivalent_to(a: Entity, b: Entity, context: Context)

  constraints:
    - "All relationship endpoints must resolve to typed entities or assertions."
    - "Confidence and evidence reliability must be within [0,1]."
    - "Distinguish observed facts, declared metadata, inferred assertions and hypotheses."
    - "Association, similarity and co-occurrence do not establish causation or identity."
    - "Do not assume transitivity or symmetry unless a predicate explicitly permits it."
    - "Model training-data relationships require documented provenance; do not infer them from output resemblance."
    - "Preserve contextual scope, timestamps, versions and uncertainty."

  inference:
    rule supported_assertion:
      if supports(e,a) and valid(e) and valid(a)
      then evidence_backed(a,e)
      confidence: "evaluate source reliability and independence"
    rule dependency_path:
      if depends_on(a,b) and depends_on(b,c) and compatible_context(a,b,c)
      then indirectly_depends_on(a,c)
      confidence: "propagate uncertainty without increasing certainty"
    rule contradiction_review:
      if contradicts(a,b) and compatible_context(a,b)
      then requires_reconciliation(a,b)
      confidence: "retain both assertions until evidence resolves conflict"

  SEMANTIC_EQUIVALENCE:
    - "Align entities only using versioned mappings and compatible context."
    - "Distinguish exact identity, contextual equivalence, approximate similarity and aliases."
    - "Preserve original identifiers and mapping provenance."

  CONFIDENCE:
    range: [0,1]
    factors: [evidence_quality, provenance, contextual_fit, uncertainty, independent_replication, model_limitations]
    policy: "Do not manufacture certainty by chaining or repeating dependent evidence."

  VALIDATION:
    - "Check entity types, predicate domain/range, referential integrity and provenance."
    - "Check context and version compatibility before inference."
    - "Flag circular support, contradictory assertions and unsupported claims."
    - "Separate deterministic logical entailment from probabilistic prediction."

  FLOW_HASH:
    algorithm: "SHA-256"
    input: "canonicalised ontology version, assertions, inference rules and provenance"
    note: "Compute hash at execution; no hash value is claimed here."

  UPDATE_MODEL:
    strategy: "versioned, provenance-preserving incremental update"
    steps: [ingest, normalise, validate, align, reconcile, infer, publish]
    rollback: true

  AI_CONSUMPTION:
    representation: "typed, context-aware knowledge graph with evidence-backed edges"
    outputs: [entity_links, relationship_triples, reasoning_traces, uncertainty_annotations, provenance_records]
    safeguards: [no_unsupported_identity_merges, no_unjustified_causal_inference, preserve_evidence_and_context, distinguish_hypotheses_from_facts]
}

END FLOW ALGEBRA