Research-to-policy chain explorer

ARI · funded project · publication · citing policy document

1,194 chains · 211 projects

1,194 reviewed chains across all seven batches. Chain ranking is provisional; project-level ranking has not been computed.

Cross-batch consistency review is still needed

Some repeated relationships have differing source assessments. Flags do not change the original recommendations.

What the flags mean
Research need, project & publicationCiting policy documentReview assessment
What do these assessments and consistency flags mean?

Priority for investigation is the supplied review’s assessment of a promising chain. It does not establish that the ARI caused the project, verify paper funding, or demonstrate policy adoption or impact.

A: ARI–paper relevance. B: ARI–policy relevance. C: project-specific chain coherence. D: substantive use of the paper in the policy document. Scores are 0–4, not probabilities. A dash means no score was supplied; zero is a recorded score.

For D: 0 = bibliography/incidental; 1 = framing; 2 = specific contribution; 3 = important argument or analysis; 4 = explicitly central use. The source label “verified substantive citation” can include D=1 background use.

Consistency review needed means the merge audit found differing assessments of a repeated relationship. It does not mean the chain is unfinished or necessarily incorrect. Citation identity needs adjudication identifies a stronger disagreement over whether the cited work is the target publication. Open a flagged chain to compare the recorded values, reasons and evidence. No judgement has been chosen over another.

No consistency flag means no issue was flagged by this comparison audit, not that the chain has been independently validated. The combined ranking standardises ordering, not scoring calibration. Original scores and recommendations are unchanged; flags do not automatically promote or demote chains.

Source platforms are not responsible departments. Parliamentary evidence can be an external submission. All quotations and PDF coverage claims are inherited from the supplied reviews and were not independently checked against full texts when this explorer was built. Complete policy PDFs are not embedded.

Filters combine. Option counts refer to the whole dataset; headline totals reflect your selection. One project or publication can belong to several different chains, so these counts should not be added together.

Dataset provenance and full-data export

CSV downloads escape formula-like text for spreadsheet safety. The JSON export preserves exact strings, all original fields, evidence, PDF-coverage records and audit comparisons. The ZIP contains the review files, compact project and ARI lookups, and editable explorer source.