Confidence Relationship Ontology

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

flow "ConfidenceRelationshipOntology" v1.0 {
  ontology:
    id: "confidence-relationship-ontology"
    parent: "ai-reasoning-relationships"
    purpose: "Represent, qualify and validate confidence attached to assertions, inferences and evidence relationships."

  types:
    Assertion: {id: string, proposition: string, context: string?, provenance: string[]}
    Evidence: {id: string, source: string, method: string?, observed_at: datetime?, reliability: number?}
    Inference: {id: string, premises: string[], conclusion: string, rule: string, assumptions: string[]}
    ConfidenceAssessment: {id: string, target: string, value: number, interpretation: enum[probability, calibrated_score, expert_judgement, heuristic], method: string, context: string?, assessed_at: datetime?, evidence: string[]}
    ConfidenceRelationship: {id: string, subject: string, predicate: enum[supports, weakens, depends_on, corroborates, contradicts, supersedes], object: string, assessment: string?, provenance: string[]}

  predicates:
    supports(evidence: Evidence, assertion: Assertion)
    weakens(evidence: Evidence, assertion: Assertion)
    confidence_of(assessment: ConfidenceAssessment, target: Assertion | Inference)
    depends_on(inference: Inference, premise: Assertion)
    corroborates(a: Evidence, b: Evidence)
    contradicts(a: Assertion, b: Assertion)
    supersedes(new: ConfidenceAssessment, old: ConfidenceAssessment)

  constraints:
    - "Confidence values must be finite and within [0, 1]."
    - "Every assessment must declare its interpretation and calculation or judgement method."
    - "A confidence score is not a probability unless calibration and probabilistic meaning are established."
    - "Correlated evidence must not be counted as independent corroboration."
    - "Absence of evidence is not automatically evidence of absence."
    - "Confidence must be scoped to a claim, context, version and relevant time."
    - "No inference chain may increase confidence merely because more steps were added."
    - "Conflicting assessments must retain their provenance rather than be silently overwritten."

  inference:
    rule evidence_support:
      if supports(e,a) and valid(e) and valid(a)
      then eligible_for_confidence_update(a,e)
      confidence: "update only using an explicitly selected and justified assessment method"
    rule premise_dependency:
      if depends_on(i,p) and confidence_of(c,p)
      then record_dependency_uncertainty(i,p,c)
      confidence: "do not assume numerical independence or multiply scores by default"
    rule contradictory_evidence:
      if contradicts(a,b) and same_context(a,b)
      then flag_confidence_conflict(a,b)
      confidence: "retain both claims pending adjudication"
    rule reassessment:
      if supersedes(new,old) and valid(new)
      then mark_previous_assessment(old,new)
      confidence: "preserve audit history"

  SEMANTIC_EQUIVALENCE:
    - "Map confidence, certainty, likelihood and reliability only when their semantics and measurement scales align."
    - "Distinguish source reliability, model confidence and proposition probability."
    - "Do not equate matching numeric scores across different calibration methods."

  CONFIDENCE:
    range: [0,1]
    factors: [evidence_quality, source_reliability, calibration, assumption_validity, context_match, evidence_dependence, temporal_relevance]
    policy: "Preserve uncertainty, calibration metadata and competing interpretations."

  VALIDATION:
    - "Check target references, data types, score bounds, evidence identifiers and provenance."
    - "Check method compatibility before aggregating or comparing confidence assessments."
    - "Flag unsupported precision, circular evidence, correlated sources and stale assessments."
    - "Require explicit treatment of missing evidence and contradictory claims."

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

  UPDATE_MODEL:
    strategy: "versioned, evidence-preserving incremental reassessment"
    steps: [ingest, normalise, validate, assess_dependencies, recalibrate_if_justified, reconcile_conflicts, recompute_inferences, publish]
    rollback: true

  AI_CONSUMPTION:
    representation: "typed evidence and inference graph with scoped confidence assessments"
    outputs: [confidence_annotations, evidence_links, inference_dependencies, conflict_flags, calibration_metadata, provenance_records]
    safeguards: [no_false_certainty, no_unjustified_probability_claims, no_double_counted_evidence, retain_conflicting_assessments]
}

END FLOW ALGEBRA