Mehmet EmrahKonyaTR
← Back to glossary

Evidence provenance

Kanıt kökeni

D4

Evidence provenance is recorded information that makes an evidence item's origin, derivation, handling, and connection to a claim inspectable.

Review status: 2026-11-26

Technical explanation

It links evidence to the entities, activities, and agents involved in producing or handling it and to the claim or decision it supports.

Conceptual boundaries

Evidence provenance is not merely citation presence; direct support, faithfulness, completeness, and sufficiency require separate evaluation.

Provider-neutral example

For an evaluation claim, a record can identify the dataset version, its source and transformation, the scoring run, the reviewer, and the exact report section that uses the result.

Limitations

A recorded provenance trail does not itself establish direct support or sufficiency for an inference.

Related concepts

Atomic claims and evidence

  1. 1.1W3C PROV-DM: The PROV Data Model
    Source
    W3C PROV-DM: The PROV Data Model
    Source role
    Authoritative source
    Exact locator
    Introduction; Sections 2.1.2-2.1.3 and 5.2-5.3
    Supported claim
    The W3C PROV model uses entities, activities, and agents and includes derivation, attribution, association, and delegation relations.
    Last verification
    Review due
    Scope limitation
    PROV is a general provenance model; it does not define a complete evidentiary standard for every AI claim.
  2. 2.1W3C PROV-Overview
    Source
    W3C PROV-Overview
    Source role
    Authoritative source
    Exact locator
    Section 1, provenance framework recommendations
    Supported claim
    W3C's PROV overview identifies identification, attribution, processing steps, reproducibility, versioning, procedures, and derivation as provenance-framework needs.
    Last verification
    Review due
    Scope limitation
    These provenance needs improve inspectability; they do not themselves prove that a claim is true or relevant.
  3. 3.1NIST, Building Evaluation Probes for Agentic AI
    Source
    NIST, Building Evaluation Probes for Agentic AI
    Source role
    Authoritative source
    Exact locator
    Overview; Objectives; Approach — factual-claim/reference-document comparison, structured audit trails, and citation-quality dimensions
    Supported claim
    NIST's evaluation-probe project uses structured audit trails to map agent decisions to supporting evidence and separately evaluates citations for faithfulness, completeness, and sufficiency.
    Last verification
    Review due
    Scope limitation
    This describes an evaluation approach and does not make every recorded provenance relation direct support for a claim.