Determinations and Relationships

Determinations and relationships describe different levels of a completed validation. They are deliberately separate.

Determination

determination is the one authoritative overall outcome for the supplied information within the supplied evidence boundary.

DeterminationCanonical meaning
supportedAll material information is supported without material contradiction.
unsupportedEvery material unit is contradicted; at unit level, contradiction exists without support.
conflictingSupport and contradiction coexist on a material unit.
uncertainPartial support, overstatement, unrelated evidence, material gaps, or mixed unit outcomes without a conflicting unit.

Every state is returned with HTTP 200 and status: "completed". unsupported does not assert universal falsity, and supported does not assert objective truth. Both are scoped to the supplied evidence.

Information-unit determinations

VI decomposes the supplied information into ordered units. Each item in information_units has its own determination, while the top-level determination remains authoritative for the complete validation.

Unit determinations make a multi-part result inspectable. They do not replace or compete with the aggregate result.

Relationships

Each relationship connects one canonical material information unit (information_unit_id) to one independent canonical evidence item (evidence_id).

RelationshipMeaning
supportsThe evidence contains the proposition with compatible scope and polarity.
contradictsThe evidence materially opposes the proposition.
partialThe material supports only part of the claim.
overstatedThe information is stronger or broader than the evidence supports.
unrelatedNo decisive relevant support or contradiction.

A relationship may contain a concise reason. It explains the classification without introducing external evidence or hidden model reasoning.

Duplicate evidence

Duplicate submissions remain in the evidence array and reference the canonical original through duplicate_of. They do not receive independent relationship evaluations and must not be presented as separate corroboration.

Canonical reason codes

The top-level reason_code is deterministic and subordinate to the determination:

DeterminationCompatible reason code or codes
supportedFULL_MATERIAL_SUPPORT
unsupportedMATERIAL_CONTRADICTION
conflictingSUPPORT_AND_CONTRADICTION
uncertainMATERIAL_EVIDENCE_GAP, PARTIAL_SUPPORT_ONLY, OVERSTATEMENT_DETECTED, NO_DECISIVE_RELATIONSHIP, or MIXED_UNIT_OUTCOMES

Clients should consume the returned determination and reason code rather than reconstructing aggregation from array ordering or relationship counts.

Example: association and causation

Information that a treatment causes improvement exceeds evidence reporting only an association: the relationship is overstated. A correction that narrows the information to what the evidence supports can yield supports; evaluate the actual returned result.