# Cascade Evidence Vocabulary (Layer 2, cross-cutting) — v1-draft # Authored in spec/. Per draft policy (cf. genomics:, advisory:), NOT registered # in spec/VOCAB_VERSIONS until v1.0 graduation. # # CHANGELOG # v1-draft.0.2 (2026-07-01) — Verdict taxonomy v2: the facet model (Workbench # grounding plan §4.1 / platform plan §4.6). The grounding outcome is now # expressed as orthogonal facets on the Assertion — evidence:direction, # evidence:basis, evidence:strength, evidence:settled, evidence:reason # (object properties over new closed enumerations DirectionValue, BasisValue, # StrengthValue, SettledValue, NeedsEvidenceReasonValue) plus # evidence:confidence (xsd:decimal, [0,1]). The FACETS are the canonical # serialized form. The flat evidence:verdict property and the VerdictValue # individuals are DEPRECATED, not removed: draft-period data and fixtures # already carry evidence:verdict, and the SHACL shapes keep a legacy branch # so that data still validates during the one-release transition. Both are # scheduled for removal at v1.0 graduation (see spec/PENDING_DOWNSTREAM_SYNC.md). # Individual reuse: evidence:Supports and evidence:Contradicts are typed as # both StanceValue (per-link) and DirectionValue (aggregate); evidence:None # is both a DirectionValue and a BasisValue; evidence:NeedsLiterature is both # a deprecated VerdictValue and a NeedsEvidenceReasonValue. Local names match # the code mirror (@cascade-workbench/contracts grounding.ts) exactly. # v1-draft.0.1 (2026-06-16) — Initial draft: Assertion / EvidenceLink / # Citation / GroundingActivity, flat VerdictValue + StanceValue enumerations, # grounding invariant. @prefix evidence: . @prefix cascade: . @prefix owl: . @prefix rdf: . @prefix rdfs: . @prefix xsd: . @prefix dct: . @prefix prov: . # ============================================================================ # Ontology Metadata # ============================================================================ a owl:Ontology ; dct:title "Cascade Protocol Evidence Vocabulary"@en ; dct:description "Layer 2 cross-cutting vocabulary for grounding assertions in cited evidence with explicit verdicts."@en ; dct:creator "Cascade Agentic Labs" ; dct:created "2026-06-16"^^xsd:date ; dct:modified "2026-07-01"^^xsd:date ; owl:versionInfo "1.0-draft.0.2" ; rdfs:comment "Generalizes the assertion -> evidence -> verdict pattern needed by Cascade Workbench (checking claims pulled from AI conversations against a patient's records and literature), the Cascade Appeal system (citing clinical evidence against policy criteria, cf. coverage:deniedClinicalContext), and PGx/genomics decision support. Reuses core provenance primitives (cascade:extractionModel, cascade:extractionConfidence, cascade:requiresUserReview, prov:wasDerivedFrom) rather than redefining them. Uses 'Assertion' (not 'Claim') to avoid collision with coverage:ClaimRecord. Draft: not in VOCAB_VERSIONS until v1.0 graduation."@en ; rdfs:seeAlso . # ============================================================================ # Classes # ============================================================================ evidence:Assertion a owl:Class ; rdfs:subClassOf prov:Entity ; rdfs:label "Assertion"@en ; rdfs:comment "A discrete, checkable statement extracted from a source (e.g. an AI conversation) or entered by the user, evaluated against evidence in the Pod and/or literature. Carries exactly one verdict and zero-or-more evidence links. The extraction of an Assertion from its source SHOULD be recorded with a cascade:AIExtractionActivity; an assertion surfaced by a general-purpose AI assistant SHOULD carry cascade:dataProvenance cascade:AIAsserted."@en ; owl:versionInfo "Added in evidence v1-draft.0.1" . evidence:EvidenceLink a owl:Class ; rdfs:label "Evidence Link"@en ; rdfs:comment "A reified link from an Assertion to one piece of evidence (a Pod record or an external Citation), carrying a stance (supports / contradicts / contextual). Modeled as a node so a single Assertion can carry multiple, differently-directed pieces of evidence. Pattern mirrors cascade:ConflictDetail."@en ; owl:versionInfo "Added in evidence v1-draft.0.1" . evidence:Citation a owl:Class ; rdfs:subClassOf prov:Entity ; rdfs:label "Citation"@en ; rdfs:comment "An external bibliographic source (e.g. a research paper) used as evidence, identified by DOI and/or PMID. When retrieved into the Pod it is a first-class resource (recommended path literature/) so an evidence chip can resolve to it exactly as it resolves to a clinical record."@en ; owl:versionInfo "Added in evidence v1-draft.0.1" . evidence:GroundingActivity a owl:Class ; rdfs:subClassOf prov:Activity ; rdfs:label "Grounding Activity"@en ; rdfs:comment "A PROV-O Activity representing the model pass that graded an Assertion against evidence and produced its verdict. Distinct from cascade:AIExtractionActivity (which extracts structured records from documents); grounding grades free-text assertions. Reuses cascade:extractionModel to record the model."@en ; owl:versionInfo "Added in evidence v1-draft.0.1" . evidence:VerdictValue a owl:Class ; owl:deprecated true ; rdfs:label "Verdict Value (deprecated)"@en ; rdfs:comment "DEPRECATED in v1-draft.0.2: the flat verdict enumeration is superseded by the facet model (evidence:direction / evidence:basis / evidence:strength / evidence:settled / evidence:reason). Retained one release so draft-period data carrying evidence:verdict still validates; clients derive the flat value from the facets in code and MUST NOT serialize it going forward. Scheduled for removal at v1.0 graduation."@en ; owl:versionInfo "Added in evidence v1-draft.0.1; deprecated in v1-draft.0.2" . evidence:StanceValue a owl:Class ; rdfs:label "Stance Value"@en ; rdfs:comment "Closed enumeration of the direction of an EvidenceLink. Instances are the named individuals below."@en ; owl:versionInfo "Added in evidence v1-draft.0.1" . # ── Facet enumeration classes (verdict taxonomy v2, grounding plan §4.1) ───── evidence:DirectionValue a owl:Class ; rdfs:label "Direction Value"@en ; rdfs:comment "Closed enumeration of what the evidence, in aggregate, says about the Assertion: Supports | Contradicts | Mixed | None. The per-link analogue is StanceValue; evidence:Supports and evidence:Contradicts are shared individuals of both."@en ; owl:versionInfo "Added in evidence v1-draft.0.2" . evidence:BasisValue a owl:Class ; rdfs:label "Basis Value"@en ; rdfs:comment "Closed enumeration of where the evidence is: Record | Literature | RecordAndLiterature | None. New bases (guideline, clinician annotation) slot in here without breaking any flat enum."@en ; owl:versionInfo "Added in evidence v1-draft.0.2" . evidence:StrengthValue a owl:Class ; rdfs:label "Strength Value"@en ; rdfs:comment "Closed enumeration of the quantity/quality of the evidence: Strong | Moderate | Weak. On basis Record, strength is a pure deterministic function of the match (guardrail G-2), never model-emitted."@en ; owl:versionInfo "Added in evidence v1-draft.0.2" . evidence:SettledValue a owl:Class ; rdfs:label "Settled Value"@en ; rdfs:comment "Closed enumeration of whether the Assertion is resolved: Settled | NeedsEvidence."@en ; owl:versionInfo "Added in evidence v1-draft.0.2" . evidence:NeedsEvidenceReasonValue a owl:Class ; rdfs:label "Needs-Evidence Reason Value"@en ; rdfs:comment "Closed enumeration of why an Assertion still needs evidence, when it does: NoRecord | NeedsLiterature | NotCheckableByNature."@en ; owl:versionInfo "Added in evidence v1-draft.0.2" . # ============================================================================ # Named Individuals (enumerations) # ============================================================================ # Deprecated flat verdicts (v1-draft.0.2): retained one release for draft-period # data; derive from the facets in code, do not serialize. Removal at v1.0. evidence:Supported a owl:NamedIndividual, evidence:VerdictValue ; owl:deprecated true ; rdfs:label "Supported (deprecated)"@en ; rdfs:comment "DEPRECATED: use evidence:settled evidence:Settled + evidence:direction evidence:Supports. The patient's records and/or literature substantiate the assertion."@en . evidence:Contradicted a owl:NamedIndividual, evidence:VerdictValue ; owl:deprecated true ; rdfs:label "Contradicted (deprecated)"@en ; rdfs:comment "DEPRECATED: use evidence:settled evidence:Settled + evidence:direction evidence:Contradicts. The patient's records and/or literature directly conflict with the assertion."@en . evidence:Unverifiable a owl:NamedIndividual, evidence:VerdictValue ; owl:deprecated true ; rdfs:label "Unverifiable (deprecated)"@en ; rdfs:comment "DEPRECATED: use evidence:settled evidence:NeedsEvidence + evidence:reason evidence:NoRecord. No evidence available in the current Pod to support or refute. NOT a judgment that the assertion is false."@en . evidence:Supports a owl:NamedIndividual, evidence:StanceValue, evidence:DirectionValue ; rdfs:label "Supports"@en ; rdfs:comment "As a StanceValue: this one link supports. As a DirectionValue (v1-draft.0.2): the evidence in aggregate supports."@en . evidence:Contradicts a owl:NamedIndividual, evidence:StanceValue, evidence:DirectionValue ; rdfs:label "Contradicts"@en ; rdfs:comment "As a StanceValue: this one link contradicts. As a DirectionValue (v1-draft.0.2): the evidence in aggregate contradicts."@en . evidence:Contextual a owl:NamedIndividual, evidence:StanceValue ; rdfs:label "Contextual"@en ; rdfs:comment "Relevant background that neither supports nor refutes."@en . # ── Facet individuals (v1-draft.0.2) ───────────────────────────────────────── evidence:Mixed a owl:NamedIndividual, evidence:DirectionValue ; rdfs:label "Mixed"@en ; rdfs:comment "The evidence points in both directions; see the individual links."@en . evidence:None a owl:NamedIndividual, evidence:DirectionValue, evidence:BasisValue ; rdfs:label "None"@en ; rdfs:comment "As a DirectionValue: the evidence says nothing. As a BasisValue: there is no evidence base. Shared individual; the property it appears on disambiguates."@en . evidence:Record a owl:NamedIndividual, evidence:BasisValue ; rdfs:label "Record"@en ; rdfs:comment "The evidence is the patient's own Pod records (deterministic path, G-2)."@en . evidence:Literature a owl:NamedIndividual, evidence:BasisValue ; rdfs:label "Literature"@en ; rdfs:comment "The evidence is published research; it has not been checked against the patient's records."@en . evidence:RecordAndLiterature a owl:NamedIndividual, evidence:BasisValue ; rdfs:label "Record and Literature"@en ; rdfs:comment "Both the patient's records and published research contribute evidence."@en . evidence:Strong a owl:NamedIndividual, evidence:StrengthValue ; rdfs:label "Strong"@en ; rdfs:comment "On the record path: exact code + value/negation match."@en . evidence:Moderate a owl:NamedIndividual, evidence:StrengthValue ; rdfs:label "Moderate"@en ; rdfs:comment "On the record path: code match with value absent or approximate."@en . evidence:Weak a owl:NamedIndividual, evidence:StrengthValue ; rdfs:label "Weak"@en ; rdfs:comment "On the record path: lexical / recall-fallback match only."@en . evidence:Settled a owl:NamedIndividual, evidence:SettledValue ; rdfs:label "Settled"@en ; rdfs:comment "The claim is resolved by the cited evidence."@en . evidence:NeedsEvidence a owl:NamedIndividual, evidence:SettledValue ; rdfs:label "Needs Evidence"@en ; rdfs:comment "The claim is not resolved; evidence:reason says why. A NeedsEvidence assertion carries direction None (a grounded direction here is facet-inconsistent and fails validation)."@en . evidence:NoRecord a owl:NamedIndividual, evidence:NeedsEvidenceReasonValue ; rdfs:label "No Record"@en ; rdfs:comment "Nothing in the patient's records speaks to this. That does not make it wrong."@en . evidence:NotCheckableByNature a owl:NamedIndividual, evidence:NeedsEvidenceReasonValue ; rdfs:label "Not Checkable By Nature"@en ; rdfs:comment "Subjective, personal, or otherwise unverifiable by its nature (instruction, opinion, tautology, meta)."@en . # evidence:NeedsLiterature doubles as a NeedsEvidenceReasonValue (below) and the # deprecated VerdictValue of the same local name; the code mirror uses the same # string for both, so one shared individual is exact. evidence:NeedsLiterature a owl:NamedIndividual, evidence:VerdictValue, evidence:NeedsEvidenceReasonValue ; rdfs:label "Needs Literature"@en ; rdfs:comment "As a NeedsEvidenceReasonValue (v1-draft.0.2): the patient's records cannot settle this; it needs published evidence. As a VerdictValue it is DEPRECATED (use evidence:settled evidence:NeedsEvidence + evidence:reason evidence:NeedsLiterature)."@en . # ============================================================================ # Properties — Assertion # ============================================================================ evidence:assertionText a owl:DatatypeProperty ; rdfs:label "assertion text"@en ; rdfs:comment "The normalized, atomic statement being evaluated."@en ; rdfs:domain evidence:Assertion ; rdfs:range xsd:string . evidence:verdict a owl:ObjectProperty ; owl:deprecated true ; rdfs:label "verdict (deprecated)"@en ; rdfs:comment "DEPRECATED in v1-draft.0.2: superseded by the facet properties (direction / basis / strength / settled / reason / confidence), which are the canonical serialized form. Retained one release so draft-period data validates; derive the flat value from the facets in code, do not serialize it. Removal at v1.0 graduation."@en ; rdfs:domain evidence:Assertion ; rdfs:range evidence:VerdictValue . # ── Facet properties (verdict taxonomy v2, v1-draft.0.2) ───────────────────── # The canonical grounding outcome. Grounding invariant (SHACL Core, see # evidence.shapes.ttl): a grounded result — settled Settled, direction one of # Supports/Contradicts/Mixed, basis not None — of EITHER basis requires at # least one evidence:hasEvidenceLink. evidence:direction a owl:ObjectProperty ; rdfs:label "direction"@en ; rdfs:comment "What the evidence, in aggregate, says about the assertion."@en ; rdfs:domain evidence:Assertion ; rdfs:range evidence:DirectionValue ; owl:versionInfo "Added in evidence v1-draft.0.2" . evidence:basis a owl:ObjectProperty ; rdfs:label "basis"@en ; rdfs:comment "Where the evidence is. Tells the UI which regime produced the result: Record is the deterministic path (G-2), Literature is model+search."@en ; rdfs:domain evidence:Assertion ; rdfs:range evidence:BasisValue ; owl:versionInfo "Added in evidence v1-draft.0.2" . evidence:strength a owl:ObjectProperty ; rdfs:label "strength"@en ; rdfs:comment "Quantity/quality of the evidence. On basis Record this is a pure deterministic function of the match (G-2), never model-emitted."@en ; rdfs:domain evidence:Assertion ; rdfs:range evidence:StrengthValue ; owl:versionInfo "Added in evidence v1-draft.0.2" . evidence:settled a owl:ObjectProperty ; rdfs:label "settled"@en ; rdfs:comment "Whether the assertion is resolved by the cited evidence."@en ; rdfs:domain evidence:Assertion ; rdfs:range evidence:SettledValue ; owl:versionInfo "Added in evidence v1-draft.0.2" . evidence:reason a owl:ObjectProperty ; rdfs:label "reason"@en ; rdfs:comment "Why the assertion still needs evidence, when settled is NeedsEvidence."@en ; rdfs:domain evidence:Assertion ; rdfs:range evidence:NeedsEvidenceReasonValue ; owl:versionInfo "Added in evidence v1-draft.0.2" . evidence:confidence a owl:DatatypeProperty ; rdfs:label "confidence"@en ; rdfs:comment "Confidence in the grounding result, [0.0, 1.0]. On basis Record this is a fixed function of match quality, never a model probability (G-2); on the literature path it may reflect model certainty. Distinct from evidence:groundingConfidence, which is the GroundingActivity's raw model confidence."@en ; rdfs:domain evidence:Assertion ; rdfs:range xsd:decimal ; owl:versionInfo "Added in evidence v1-draft.0.2" . evidence:rationale a owl:DatatypeProperty ; rdfs:label "rationale"@en ; rdfs:comment "One-line plain-language explanation of the verdict."@en ; rdfs:domain evidence:Assertion ; rdfs:range xsd:string . evidence:disconfirming a owl:DatatypeProperty ; rdfs:label "disconfirming consideration"@en ; rdfs:comment "A 'what would argue against this' consideration. Multiple permitted. Required by product design for any material assertion (anti-confirmation-bias)."@en ; rdfs:domain evidence:Assertion ; rdfs:range xsd:string . evidence:hasEvidenceLink a owl:ObjectProperty ; rdfs:label "has evidence link"@en ; rdfs:domain evidence:Assertion ; rdfs:range evidence:EvidenceLink . evidence:groundedBy a owl:ObjectProperty ; rdfs:subPropertyOf prov:wasGeneratedBy ; rdfs:label "grounded by"@en ; rdfs:comment "Links an Assertion to the GroundingActivity that produced its verdict."@en ; rdfs:domain evidence:Assertion ; rdfs:range evidence:GroundingActivity . evidence:assertedAt a owl:DatatypeProperty ; rdfs:label "asserted at"@en ; rdfs:domain evidence:Assertion ; rdfs:range xsd:dateTime . # Reuse (NOT redefined here): cascade:extractionModel, cascade:extractionConfidence, # and cascade:requiresUserReview carry the extraction provenance of an Assertion. # ============================================================================ # Properties — EvidenceLink # ============================================================================ evidence:evidenceStance a owl:ObjectProperty ; rdfs:label "evidence stance"@en ; rdfs:domain evidence:EvidenceLink ; rdfs:range evidence:StanceValue . evidence:citesRecord a owl:ObjectProperty ; rdfs:subPropertyOf prov:wasDerivedFrom ; rdfs:label "cites record"@en ; rdfs:comment "Points to a Pod record (clinical:, health:, pots:, genomics: ...) used as evidence. Range intentionally open (rdfs:Resource) for cross-namespace linking, mirroring coverage:deniedClinicalContext."@en ; rdfs:domain evidence:EvidenceLink ; rdfs:range rdfs:Resource . evidence:citesSource a owl:ObjectProperty ; rdfs:label "cites source"@en ; rdfs:comment "Points to an external Citation (research paper) used as evidence."@en ; rdfs:domain evidence:EvidenceLink ; rdfs:range evidence:Citation . evidence:evidenceWeight a owl:DatatypeProperty ; rdfs:label "evidence weight"@en ; rdfs:comment "Optional 0.0-1.0 strength of this evidence link toward the verdict."@en ; rdfs:domain evidence:EvidenceLink ; rdfs:range xsd:decimal . # ============================================================================ # Properties — Citation # ============================================================================ evidence:doi a owl:DatatypeProperty ; rdfs:label "DOI"@en ; rdfs:domain evidence:Citation ; rdfs:range xsd:string . evidence:pmid a owl:DatatypeProperty ; rdfs:label "PMID"@en ; rdfs:domain evidence:Citation ; rdfs:range xsd:string . evidence:citationTitle a owl:DatatypeProperty ; rdfs:label "title"@en ; rdfs:domain evidence:Citation ; rdfs:range xsd:string . evidence:citationYear a owl:DatatypeProperty ; rdfs:label "year"@en ; rdfs:domain evidence:Citation ; rdfs:range xsd:gYear . evidence:citationSourceName a owl:DatatypeProperty ; rdfs:label "source name"@en ; rdfs:comment "Retrieval source, e.g. 'PubMed', 'Europe PMC'."@en ; rdfs:domain evidence:Citation ; rdfs:range xsd:string . evidence:retrievedAt a owl:DatatypeProperty ; rdfs:label "retrieved at"@en ; rdfs:domain evidence:Citation ; rdfs:range xsd:dateTime . # ============================================================================ # Properties — GroundingActivity # ============================================================================ evidence:groundingConfidence a owl:DatatypeProperty ; rdfs:label "grounding confidence"@en ; rdfs:comment "Model confidence in the produced verdict, [0.0, 1.0]."@en ; rdfs:domain evidence:GroundingActivity ; rdfs:range xsd:decimal . # Reuses cascade:extractionModel to record which model performed the grounding.