Voting Records
There are two calculation engines in this product. One reasons over layered context to answer a single decision. This is the other one — the one that hunts.
It goes and finds the public record: which enacted instrument carries a problem, which vehicle it rode inside, and how every member voted at every stage of it.
What it produces
For each problem in scope:
- The enacted-law record — which instrument actually carries the harm.
- The roll-call record — a table of how members voted, at each stage.
- Stage weighting — because not all stages are equal, and a vote in committee is not a vote on the floor.
- A per-problem and an overall figure that fills the voting-record block on an elected politician's page.
- Gaps — everything the run could not establish, listed rather than silently averaged over.
The hard line this engine holds
A public fact and the movement's reading of it are different things, stored in different fields.
How somebody voted is a fact with a citation. Whether that vote advanced or resisted the harm is the movement's judgment — one explicit, reviewable field, set deliberately, and never allowed to hide inside what looks like raw data.
Keeping them separate is what lets somebody accept our data while rejecting our reading, which is a thing a reader should always be able to do.
What a run may write
Bounded on purpose, and every run leaves an output record naming what it touched. A calculation that can quietly rewrite the record it is calculating from is not evidence.
Pages in this area
- Where the data comes from — sources and provenance.
- What counts as a vote — votes, authorship, stages, vehicles.
- Blocked from a vote — the act with the biggest effect and the smallest trace.
Where to go next
- The Flamingo Award — the area before this one.
- Qualifications — the area after this one.
- Open the app — do this for real, on your own seats.