Data & provenance

Provenance system

A dataset is a directory with a manifest.json (dataset kind — synthetic, pilot, or real; a SHA-256 of the trial CSV; the git commit and experiment version that generated it) plus trials.csv. The analysis package refuses to load a dataset whose file does not match its manifest's checksum. Every downstream artefact — a results JSON file, a figure — embeds the same information plus the analysis script and version that produced it, and that block is rendered under every chart on this site.

What exists today

Only synthetic datasets, generated by an archetype-based participant simulator described in methods. There is no pilot or real dataset. When real collection begins, this page and every chart will update automatically — the code does not distinguish data sources except by the label in the manifest.

Exports

Research exports are served from /api/export (CSV or JSON), gated by a bearer token that is never exposed to the browser. Personally identifying information is never collected: no names, emails, or IP addresses; two demographic fields are optional; an external recruitment-platform id, if supplied, is stored only as a salted hash.

Trial record schema (35 columns)