Kept Word · method
How a verdict gets made
Said, then did. We pin concrete promises to the record — and we never claim more than the record shows.
Verdicts are evidence-bound
Every Kept, Broken, or Partial label rests on a sourced evidence row. No evidence, no verdict — the promise falls back to PENDING and makes no claim. You are never asked to trust a badge on its own.
Why “evidence-bound”?
The dissent log is the defamation shield
Any counter-evidence sits right next to the verdict in the dissent log. If an official narrowed a cap in a separate bill, restored funds in conference, or held tele-town-halls the calendar missed, that goes on the record beside the label. This keeps the page fair to the official — and it is the same record that protects Kept Word: we publish the verdict and what cuts against it, every time.
No scoring, no editorial ranking
We don't grade officials or sort them into winners and losers. Kept Word puts concrete promises and the underlying records side by side and lets you read them. The compare view tallies only evidence-backed verdicts — it is a side-by-side, not a scoreboard.
Sources
- Congress.gov — federal bills and roll-call votes. The home page already surfaces live records pulled directly from the official Congress.gov API. We reflect the source; we never auto-derive promise-keeping from it.
- OpenStates — state legislators, bills, and votes for officials below the federal level.
On the record now
Curated, evidence-bound records currently tracked (2 officials in beta):
Kept Word is a free, grant-funded prototype from America's Future. Curated records are seed/illustrative in beta and persist to an accountability database when one is configured (graceful fallback to seed when not). Verdicts only render with sourced evidence plus a dissent log; promise-keeping is never auto-derived. Live congressional records come straight from the official Congress.gov API. No "you are safe" assurances, no scoring — every value shows its provenance.