Memory Relationship Ontology

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

flow "MemoryRelationshipOntology" v1.0 {
  ontology:
    id: "memory-relationship-ontology"
    parent: "ai-reasoning-relationships"
    purpose: "Represent evidence-backed links among AI memory records, entities, contexts, events and reasoning outcomes without conflating recall with factual truth."

  types:
    MemoryRecord: {id: string, kind: enum[episodic, semantic, procedural, working, external], content_ref: string, created_at: datetime, valid_from: datetime?, valid_until: datetime?, provenance: string, sensitivity: string?}
    Entity: {id: string, type: string, label: string}
    Context: {id: string, scope: string, timestamp: datetime?}
    Evidence: {id: string, source_ref: string, collected_at: datetime, reliability: number?}
    ReasoningStep: {id: string, rule_id: string, input_refs: string[], output_ref: string}
    Relationship: {id: string, subject: string, predicate: string, object: string, context_ref: string?, confidence: number, evidence_refs: string[], provenance: string}

  predicates:
    refers_to(memory: MemoryRecord, entity: Entity)
    recorded_in(memory: MemoryRecord, context: Context)
    derived_from(memory: MemoryRecord, source: Evidence)
    updates(newer: MemoryRecord, older: MemoryRecord)
    contradicts(a: MemoryRecord, b: MemoryRecord)
    supports(memory: MemoryRecord, step: ReasoningStep)
    retrieved_for(memory: MemoryRecord, context: Context)
    semantically_related(a: MemoryRecord, b: MemoryRecord)

  constraints:
    - "All relationship endpoints must resolve to typed identifiers."
    - "Confidence values must lie within [0, 1]."
    - "Retrieval relevance is not evidence of truth or consent to reuse."
    - "Distinguish recorded, inferred, corrected, superseded and retracted memories."
    - "Do not infer identity from similar labels or overlapping descriptions alone."
    - "Preserve time, context, source and access restrictions on every derived link."
    - "Updates must not silently erase historical provenance or conflict records."
    - "Do not assume semantically_related, supports or updates is transitive."

  inference:
    rule contextual_retrieval:
      if recorded_in(m,c) and context_matches(c,query_context) and authorised(m,query_context)
      then candidate_for_retrieval(m,query_context)
      confidence: "relevance must be separately evaluated from factual confidence"
    rule provenance_trace:
      if derived_from(m,e) and supports(m,s)
      then trace_evidence(s,e)
      confidence: "retain original source reliability and inference assumptions"
    rule memory_revision:
      if updates(newer,older) and valid_revision(newer,older)
      then superseded_by(older,newer)
      confidence: "preserve both versions and their provenance"
    rule contradiction_review:
      if contradicts(a,b) and same_scope(a,b)
      then requires_reconciliation(a,b)
      confidence: "do not automatically discard either record"

  SEMANTIC_EQUIVALENCE:
    - "Normalise vocabulary using versioned concept mappings."
    - "Keep alias, similarity, reference, entailment and identity distinct."
    - "Require compatible contexts and evidence before merging memory records."

  CONFIDENCE:
    range: [0,1]
    factors: [source_reliability, temporal_relevance, contextual_match, evidence_quality, contradiction_status, inference_depth]
    policy: "Calibrate confidence independently of retrieval score; do not manufacture certainty through repeated recall."

  VALIDATION:
    - "Check entity resolution, reference integrity, context, timestamps and provenance."
    - "Flag stale, superseded, unsupported and contradictory memory relationships."
    - "Enforce authorisation, privacy restrictions and retention policies before retrieval or export."
    - "Retain unresolved interpretations when evidence is insufficient."

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

  UPDATE_MODEL:
    strategy: "versioned, provenance-preserving incremental reconciliation"
    steps: [ingest, classify, normalise, validate, authorise, compare, reconcile, update_links, recompute_inferences, publish]
    rollback: true

  AI_CONSUMPTION:
    representation: "typed memory relationship graph with temporal, contextual and evidence annotations"
    outputs: [memory_links, entity_references, retrieval_candidates, provenance_traces, revision_history, uncertainty_annotations]
    safeguards: [privacy_by_design, no_unsupported_identity_merges, no_recall_as_truth, respect_retention_and_access, preserve_conflicts]
}

END FLOW ALGEBRA