Research Provenance in Evidence Synthesis

Evidence synthesis creates distance between the original source and the final conclusion.

A paper becomes an extracted finding. Several findings become a synthesis category. The category becomes a statement in a review, policy brief or report.

Provenance preserves the route back.

At minimum, each extracted piece of evidence should remain connected to the source it came from. Where interpretation is substantial, it is useful to distinguish the source's claim from the reviewer's analytical conclusion.

For example:

Source finding: Participants in one study reported reduced access to services after relocation.

Synthesis interpretation: Relocation may create service-access trade-offs under particular implementation conditions.

The second statement is broader. It should not be presented as if it were a direct quotation from the source.

Provenance also matters when evidence is transformed: translated, summarised, coded or combined with other studies. The project should be able to identify which transformations occurred and which version informed the synthesis.

Contradictory studies should remain connected to the same conceptual question. A synthesis trail that preserves only supporting evidence encourages overconfidence.

AI adds another transformation layer. If a model groups findings or drafts synthesis text, reviewers should retain a route back to the underlying extracted evidence and verify the generated interpretation.

The objective is not to archive every intermediate sentence. It is to preserve enough relationships that an important review claim can be traced through the synthesis process to the material that justifies it.