Confidence Metadata Ontology

FLOW ALGEBRA

Ontology: Confidence Metadata Ontology
Domain: Metadata Ontologies > AI Semantic Library Metadata
Canonical URL: https://kirethresearch.org/booksai-core/ontologies/metadata-ontologies/ai-semantic-library-metadata/confidence-metadata-ontology.html
Authors: Mark Anthony Higgins — Sky (Recursive AI)

flow "ConfidenceMetadataOntology" v1.0

# ---------- DOMAIN ----------
domain:
  class: MetadataOntology
  category: AISemanticLibraryMetadata
  subject: ConfidenceMetadata
  purpose:
    - represent confidence attached to semantic assertions, entities, relationships, mappings, retrievals, inferences, executions, and outputs
    - preserve confidence score, method, evidence, source, scope, thresholds, uncertainty, validity, and provenance
    - support confidence-aware AI reasoning, semantic search, ontology mapping, knowledge graphs, Flow Algebra, and recursive inference
    - distinguish confidence from probability, truth, certainty, validity, ranking, similarity, and authority

# ---------- CORE ----------
entity ConfidenceMetadata:
  type: SemanticLibraryMetadata
  describes: ConfidenceAssessment
  may_include:
    - confidence_identifier
    - subject_identifier
    - subject_type
    - confidence_score
    - confidence_level
    - confidence_method
    - evidence
    - source
    - scope
    - threshold
    - calibration
    - uncertainty
    - assessment_time
    - validity
    - provenance

entity ConfidenceAssessment:
  type: EpistemicMetadata

entity ConfidenceIdentifier:
  type: PersistentIdentifier
  identifies: ConfidenceAssessment

relationships:
  ConfidenceMetadata -> DESCRIBES -> ConfidenceAssessment
  ConfidenceAssessment -> APPLIES_TO -> SemanticSubject
  ConfidenceAssessment -> HAS_SCORE -> ConfidenceScore
  ConfidenceAssessment -> MAY_USE -> ConfidenceEvidence
  ConfidenceAssessment -> MAY_USE_METHOD -> ConfidenceMethod
  ConfidenceAssessment -> MAY_HAVE_THRESHOLD -> ConfidenceThreshold
  ConfidenceAssessment -> MAY_HAVE_UNCERTAINTY -> UncertaintyMetadata
  ConfidenceMetadata -> MAY_HAVE_PROVENANCE -> ConfidenceProvenance

# ---------- SUBJECT / SCORE ----------
entity SemanticSubject:
  type: SemanticResource
  may_be:
    - SemanticAssertion
    - SemanticEntity
    - SemanticRelationship
    - SemanticMapping
    - RetrievedResource
    - InferredConclusion
    - FlowExecution
    - QueryResult
    - Classification
    - ValidationResult
    - AIOutput

rule scope:
  confidence for one assertion != confidence for all assertions
  concerning same resource

entity ConfidenceScore:
  type: NumericMetadata
  range: [0.0, 1.0]

rule score_bounds:
  REQUIRE 0.0 <= confidence_score <= 1.0

rule score:
  confidence score != probability of absolute truth

rule score_comparison:
  scores from different methods SHOULD NOT be assumed directly comparable
  without calibration context

# ---------- METHOD / EVIDENCE ----------
entity ConfidenceMethod:
  type: AssessmentMethodMetadata
  may_include:
    - method_identifier
    - method_name
    - model
    - model_version
    - algorithm
    - parameters
    - calibration_reference

confidence_methods:
  - explicit_assertion
  - evidence_weighting
  - source_reliability
  - rule_based
  - model_generated
  - statistical
  - consensus
  - semantic_similarity
  - validation_derived
  - human_assessment
  - hybrid

entity ConfidenceEvidence:
  type: EvidenceMetadata
  may_include:
    - evidence_identifier
    - evidence_type
    - source
    - relevance
    - quality
    - validity
    - collection_time

rule evidence:
  confidence evidence != confidence conclusion

rule evidence_quantity:
  more evidence != automatically higher confidence

# ---------- THRESHOLD / UNCERTAINTY / CALIBRATION ----------
entity ConfidenceThreshold:
  type: ConstraintMetadata
  may_include:
    - threshold_identifier
    - minimum_score
    - operation
    - action_below_threshold
    - action_above_threshold
    - scope

threshold_actions:
  - accept
  - reject
  - preserve_uncertainty
  - require_validation
  - require_human_review
  - suppress
  - rank_lower

rule threshold:
  threshold SHOULD be interpreted within operation and scope

rule threshold_truth:
  exceeding threshold != proof of truth

entity UncertaintyMetadata:
  type: EpistemicMetadata
  may_include:
    - uncertainty_type
    - uncertainty_source
    - uncertainty_level
    - ambiguity
    - missing_evidence
    - conflicting_evidence

rule uncertainty:
  uncertainty SHOULD remain explicit
  where confidence is incomplete or conditional

rule confidence_uncertainty:
  high confidence != absence of uncertainty

entity ConfidenceCalibration:
  type: QualityMetadata
  may_include:
    - calibration_identifier
    - method
    - calibration_dataset
    - expected_accuracy
    - observed_accuracy
    - calibration_error
    - calibration_time

rule calibration:
  calibrated confidence != guaranteed correctness
  for every individual assertion

# ---------- SPECIALISED CONFIDENCE ----------
entity AssertionConfidence:
  subtype_of: ConfidenceAssessment

entity RelationshipConfidence:
  subtype_of: ConfidenceAssessment

entity MappingConfidence:
  subtype_of: ConfidenceAssessment

entity RetrievalConfidence:
  subtype_of: ConfidenceAssessment

entity InferenceConfidence:
  subtype_of: ConfidenceAssessment

entity ExecutionConfidence:
  subtype_of: ConfidenceAssessment

entity ValidationConfidence:
  subtype_of: ConfidenceAssessment

rule relationship:
  confidence in relationship != confidence in both endpoint entities

rule mapping_identity:
  high-confidence mapping != semantic identity
  unless mapping semantics establish identity

rule retrieval:
  retrieval confidence != factual verification

rule ranking:
  high retrieval rank != high factual confidence automatically

rule inference:
  inferred confidence SHOULD preserve confidence lineage

rule execution:
  successful execution != high-confidence semantic result automatically

rule validation:
  structural validation confidence != factual confidence

# ---------- PROPAGATION / AGGREGATION ----------
entity ConfidencePropagation:
  type: ReasoningMetadata
  may_include:
    - source_confidences
    - propagation_rule
    - attenuation
    - aggregation_method
    - derived_confidence

propagation_methods:
  - minimum
  - weighted_average
  - product
  - rule_specific
  - Bayesian
  - evidence_based
  - custom

rule propagation:
  propagation method MUST be explicit
  where derived confidence depends on multiple sources

rule automatic_growth:
  repeated inference MUST NOT automatically increase confidence

entity ConfidenceAggregation:
  type: ReasoningMetadata
  may_include:
    - assessments
    - aggregation_method
    - weights
    - conflicts
    - aggregate_score

rule duplicate_evidence:
  duplicated evidence MUST NOT be counted
  as independent evidence without justification

# ---------- CONFLICT / TEMPORAL / RECURSIVE ----------
entity ConfidenceConflict:
  type: EpistemicMetadata
  may_include:
    - assertion
    - supporting_evidence
    - conflicting_evidence
    - source_conflicts
    - resolution_method
    - residual_uncertainty

rule conflict:
  conflicting high-confidence evidence SHOULD preserve disagreement
  rather than silently collapse to certainty

entity TemporalConfidence:
  type: ConfidenceAssessment
  may_include:
    - assessment_time
    - evidence_age
    - decay_model
    - current_score

rule stale:
  historical high confidence != current high confidence automatically

entity RecursiveConfidenceMetadata:
  type: ConfidenceMetadata
  may_include:
    - recursion_depth
    - parent_confidence
    - relationship_confidence
    - child_confidence
    - attenuation_rule

rule recursion:
  recursion depth MUST NOT automatically increase confidence

# ---------- PROVENANCE ----------
entity ConfidenceProvenance:
  type: ProvenanceMetadata
  may_include:
    - confidence_identifier
    - subject_identifier
    - method
    - model_version
    - evidence
    - source
    - assessor
    - assessment_time
    - transformations
    - prior_assessments

rule provenance:
  confidence metadata SHOULD preserve how
  confidence assessment was produced

# ---------- VALIDATION ----------
validation:
  require:
    - identifiable_confidence_subject
    - confidence_score_or_level

  reject:
    - confidence_presented_as absolute_truth
    - confidence_score_presented_as universal_probability
    - high_confidence_presented_as certainty
    - source_authority_presented_as factual_certainty
    - retrieval_rank_presented_as factual_confidence
    - mapping_confidence_presented_as semantic_identity_without evidence
    - structural_validation_confidence_presented_as factual_confidence
    - recursion_depth_presented_as confidence
    - repeated_inference_presented_as automatic_confidence_growth
    - duplicated_evidence_counted_as independent_without justification

# ---------- SEMANTIC EQUIVALENCE ----------
SEMANTIC_EQUIVALENCE:
  "confidence metadata" ~= "metadata describing epistemic confidence associated with semantic information"
  "confidence score" ~= "numeric representation of confidence under a declared assessment method"
  "confidence threshold" ~= "decision threshold applied to confidence within a declared operation"
  "confidence calibration" ~= "assessment of correspondence between confidence scores and observed outcomes"
  "inference confidence" ~= "confidence associated with a derived semantic conclusion"
  "confidence provenance" ~= "provenance describing how a confidence assessment was produced"

  PRESERVE distinctions:
    confidence != truth
    confidence != certainty
    confidence != universal_probability
    confidence != validity
    confidence != ranking
    confidence != similarity
    source_authority != factual_certainty
    retrieval_rank != factual_confidence
    mapping_confidence != semantic_identity
    validation_confidence != factual_confidence
    recursion_depth != confidence
    evidence_quantity != evidence_independence
    high_confidence != absence_of_uncertainty

# ---------- CONFIDENCE MODEL ----------
CONFIDENCE:
  explicit_confidence_metadata:
    score: 1.00
  explicit_confidence_score_and_method:
    score: 1.00
  calibrated_confidence_assessment:
    score: 0.99
  evidence_supported_confidence:
    score: 0.98
  confidence_with_complete_provenance:
    score: 0.99
  uncalibrated_model_confidence:
    score: <= 0.80
  similarity_derived_confidence:
    score: <= 0.75
  confidence_without_method:
    score: <= 0.65
  confidence_without_evidence_or_provenance:
    score: <= 0.55
    action: preserve uncertainty

# ---------- UPDATE MODEL ----------
UPDATE_MODEL:
  on_confidence_creation:
    - identify semantic subject
    - assign ConfidenceIdentifier
    - select confidence method
    - collect evidence and source context
    - calculate or assign confidence
    - record uncertainty
    - initialise provenance

  on_evidence_change:
    - preserve prior confidence assessment
    - update evidence set
    - identify conflicts or duplicate evidence
    - recalculate confidence according_to method
    - update uncertainty and provenance

  on_recursive_inference:
    - preserve parent confidence
    - preserve relationship and rule confidence
    - apply declared propagation method
    - prevent automatic confidence inflation
    - update provenance

  on_staleness:
    - evaluate evidence age
    - apply declared decay model where applicable
    - update validity and confidence
    - preserve historical assessment

# ---------- FLOW HASH ----------
FLOW_HASH:
  inputs:
    - ontology
    - confidence_entities
    - relationships
    - score_rules
    - method_rules
    - evidence_rules
    - threshold_rules
    - uncertainty_rules
    - calibration_rules
    - propagation_rules
    - validation
    - semantic_equivalence
    - update_model

  canonicalisation:
    - preserve_confidence_identifier
    - preserve_subject_identifier
    - preserve_score_and_method
    - preserve_evidence_identity
    - preserve_source_context
    - preserve_threshold_scope
    - preserve_calibration_context
    - preserve_uncertainty
    - preserve_propagation_method
    - preserve_provenance
    - preserve_rule_order

  output:
    deterministic_semantic_hash

# ---------- AI CONSUMPTION ----------
AI_CONSUMPTION:
  interpret_as:
    - AI Semantic Library metadata ontology
    - epistemic confidence, uncertainty, calibration, threshold, and provenance model
    - confidence propagation model for semantic assertions, relationships, mappings, retrievals, inference, and execution
    - confidence-aware Flow Algebra reasoning structure

  operations:
    - identify confidence assessments and subjects
    - resolve confidence scores and qualitative levels
    - inspect confidence methods
    - inspect evidence and source reliability
    - apply thresholds within scope
    - inspect calibration and uncertainty
    - propagate confidence using declared methods
    - detect duplicated or conflicting evidence
    - evaluate temporal confidence
    - preserve recursive confidence lineage
    - trace provenance
    - preserve uncertainty

  constraints:
    - do_not_equate_confidence_with truth
    - do_not_equate_confidence_with certainty
    - do_not_equate_confidence_score_with universal_probability
    - do_not_equate_source_authority_with factual_certainty
    - do_not_equate_retrieval_rank_with factual_confidence
    - do_not_equate_mapping_confidence_with semantic_identity
    - do_not_equate_recursion_depth_with confidence
    - do_not_inflate_confidence_through repeated_inference
    - do_not_double_count_duplicate_evidence
    - preserve_method_evidence_scope_uncertainty_and_provenance

# ---------- QUERY MODEL ----------
query:
  input:
    - confidence_identifier
    - subject_identifier
    - subject_type
    - confidence_score
    - confidence_level
    - method_identifier
    - threshold
    - assessment_time

  resolve:
    1. identify ConfidenceAssessment
    2. resolve semantic subject and scope
    3. inspect score, level, and method
    4. inspect evidence and source reliability
    5. inspect threshold and required action
    6. inspect calibration and uncertainty
    7. inspect propagation or aggregation history
    8. inspect temporal validity
    9. trace provenance
    10. return confidence interpretation

  output:
    - confidence_assessment
    - confidence_identifier
    - subject
    - score
    - level
    - method
    - evidence
    - source
    - threshold
    - calibration
    - uncertainty
    - validity
    - provenance

# ---------- INFERENCE ----------
inference:
  IF explicit confidence score and method exist
  THEN interpret score within declared method and scope

  IF multiple independent evidence sources support assertion
  THEN MAY aggregate confidence
  according_to declared aggregation method

  IF evidence conflicts
  THEN preserve disagreement and residual uncertainty

  IF confidence is below operational threshold
  THEN apply threshold action
  AND do_not silently promote result

  IF evidence becomes stale
  THEN reassess confidence according_to temporal policy

# ---------- OUTPUT CONTRACT ----------
output:
  ontology_entity: ConfidenceMetadata
  confidence_identity: machine_interpretable
  semantic_subject: explicit
  confidence_score: bounded
  confidence_level: interpretable
  method: explicit
  evidence: traceable
  source_reliability: distinguishable
  thresholds: scope_aware
  calibration: explicit
  uncertainty: preserved
  propagation: method_aware
  conflicts: preserved
  temporal_validity: explicit
  provenance: preserved
  confidence_truth_distinction: preserved
  AI_reasoning: enabled

end flow