PROV-DM: The PROV Data Model

Source: W3C Author: W3C Date: 2013

Core idea

PROV-DM models provenance with first-class entities, activities, agents, and relations such as generation, usage, derivation, attribution, association, revision, and invalidation.

Key claims

• Provenance should be explicit graph structure, not an afterthought hidden in logs.
• Revisions and invalidations deserve first-class representation.
• Provenance can itself be bundled and attributed, supporting provenance-of-provenance.

Harness takeaway

An observed-goals ledger should be built on a provenance graph so facts, receipts, hypotheses, revisions, and outcomes remain inspectable and queryable instead of being flattened into summaries or scores.