How JustMeasure Works

JustMeasure is a political accountability tool that gathers public data from official congressional voting records and bill metadata. Candidate rankings come from those votes.

Bill summaries and tags are drafted to classify:

• what a bill is about
• what the vote appears to do in practical terms
• how the bill may relate to specific biblical and moral concerns already defined by the project

That step uses Claude Haiku 5.5 (anthropic/claude-haiku-5.5), prompt bill-enrichment-v6.1, temperature 0.1. The model sees the bill id, title, summary, policy area, latest action, chamber, and source URL, plus tags already on the bill. It does not see the sponsor or the sponsor's party. Appropriations, war funding, omnibus bills, a family-policy tag paired with life or child protection, recommit motions, continuing appropriations, and titles that say “for other purposes” beside appropriations or an authorization go to Claude Sonnet 5.5 (anthropic/claude-sonnet-5.5). So do mixed or unclear readings, a bill marked needs review, and any restraint-lane command the first model flags.

A bill marked needs review does not score until a person reviews it. On the restraint lane, a low-confidence result does not score. An ordinary tag is not held back only because confidence is low. The moral rules are human-written. Verses on ordinary tags are human-authored. Restraint citations have to be on a fixed list. The model does not create the moral standard.

Roll calls come from the House Clerk and Senate LIS. Bills and CRS summaries come from the Congress.gov API. Names and offices come from FEC bulk files. Sitting members are checked against the House Clerk member list and the Senate contact list.

Ballot briefs are separate from candidate rankings. They grade the Yes and No options on a public measure with the same Scripture standard and the same five restraint tests used for bills. No means current law. The leans are Yes, No, Conditional Yes, Conditional No, and Split. Claude Haiku 5.5 proposes the lanes. A person reviews them. A brief moves from draft, to draft reviewed, to published. These briefs are counsel to conscience (Westminster Confession of Faith 20.2). They do not score voters and they do not enter legislator rankings. See Ballot briefs.

My Ballot is the voting-booth cheat sheet. It opens on the page you are already on. Senate races are Class II seats, whose terms end January 3, 2027 (Senate contact list, retrieved October 19, 2026; Class II), plus two sourced specials: Florida, Ashley Moody (Florida Division of Elections), and Ohio, Jon Husted, unexpired term ending January 3, 2029 (Ohio Secretary of State Directive 2026-18). Every U.S. House seat is on the ballot (U.S. Const. art. I, §2). Only those incumbents are graded. Challengers and newcomers are not named and are not scored.

Bills affect rankings when they are mapped into reviewed scoring categories and marked ready for scoring. A bill marked needs review does not score until a person reviews it. Presidential nominations (ids that start with pn-) and House simple resolutions (hres-) are not graded. A Congressional Review Act disapproval does not score unless a reviewed enrichment marks it ready to score.

From there, JustMeasure ranks candidates using a three-tier system.

Tier 1: Hard disqualifiers

Tier 1 marks a candidate Disqualified when either of these is true:

• a yea for a non-bundled bill tagged abortion, child sexual corruption, bloodguilt, or open corruption, on final passage or an amendment at full weight, or on cloture, a motion to proceed, or discharge when the direction is known
• a nay on cloture, the motion to proceed, or final passage of a clean bill that defends life or protects children, when that member has no yea on final passage or an amendment of the same bill

A yea on cloture is not substantive support. A later substantive yea keeps a cloture nay at Tier 2. Clean means the known content includes that protection and no unrelated major tag. Family policy is always a companion. Abuse accountability is a companion only together with child protection. Unknown content is not clean. A reviewer-verified overreach provision, a reviewed Tier 2 cap, or a bundled bill blocks Tier 1. The event stores 100. It is not reduced by the vote-type multiplier. Sponsoring such a bill can disqualify. Cosponsoring scores at half weight and is not Tier 1 by itself.

Tier 2: Grave compromise

Tier 2 looks for patterns of grave compromise, such as repeated support for:

• war funding or foreign entanglement
• debt expansion and inflationary spending
• omnibus packaging that hides accountability
• surveillance or coercive state power

Those patterns are scored as a lane profile, not as a running total. Each lane's rate is the member's residual divided by the exposure on bills in that lane, times the heaviest penalty weight that fired in that lane in the member's chamber. The lanes are summed. Repeating the same compromise inside one lane does not pile the weights up. With no Tier 1 failure, a lane-profile score of 50 or more is Unfit. A score of 25 or more, and below 50, is Compromised.

Surveillance still scores: a yea is a Tier 2 penalty of 15, and a nay is a credit of 8. Speech protection, religious liberty, and due process score at 8 either way. The restraint lane (Westminster Confession of Faith 20.4) tests both the bill and current law with five symmetric tests. Only a Scripture command can cost points, both models must agree, and low confidence on that lane does not score. The lane weight would be 15 and a credit 8, but the reviewed outcome stored no penalties across 417 bills, so the lane has changed no score. The notes are not on bill pages yet.

Three-vote minimum. Tier 2 measures a pattern, and a pattern needs more than one witness: "a matter must be established by the testimony of two or three witnesses" (Deuteronomy 19:15). Each Tier 2 lane (life, children, honest money, war and peace, and so on) gets its full weight only when a member has cast at least 3 distinct scored roll-call votes in it. With fewer, the lane counts proportionally: 1 vote counts at 1/3 weight, 2 votes at 2/3. The candidate breakdown shows this, for example "life 16.7 (1 of 3 required votes, so 1/3 weight; 100% of 50)". Tier 1 disqualifiers are not affected.

Overreach provisions. Some protective bills also hand the government a power that Scripture does not give it, such as a deadline to take down content with no appeal, or seizing property without a conviction. A member can vote against such a bill without voting against the protection. When a reviewer has checked the bill text and quoted the provision, the bill is treated like a bundled bill. A nay on it is never a Tier 1 disqualifier, and its Tier 2 penalties count at half weight. The provision has to be a means of enforcement, never the bill's ban on the harm itself. For example, S. 146 (TAKE IT DOWN Act) is flagged for its 48-hour takedown rule with no counter-notice.

Contested credits and tier-2 nays at 25. When a member votes against the penalized side on a genuinely contested final vote (final passage, a conference report, or cloture when the bill had neither, with at least 10% of yea and nay votes on the losing side), that vote earns a credit in the same lane. The test uses vote margins only, never party. A credit can offset at most half of a lane's Tier 2 penalties on other bills, never penalties from bundled bills, and never counts toward the three-vote minimum. Credits never affect Tier 1. Separately, a bill a reviewer has capped at Tier 2 counts at half weight, so a nay on a 50-weight protective bill counts 25. A bundled bill keeps its own half weight; the two are never combined.

Tier 3: Positive fitness

Tier 3 looks for positive fitness, meaning evidence that a candidate:

• actively defends life
• protects children
• supports accountability and truth-telling
• resists unjust war and favors de-escalation
• shows stronger stewardship and independence from corrupt systems

Candidates who clear Tier 1 and Tier 2 can then be ranked more positively based on those strengths. In other words, positive points do not erase direct support for grave evil.

The final outcomes are typically:

• Disqualified
• Unfit
• Compromised
• Qualified: below 25, no Tier 1 failure, enough votes, and either a scored negative record or a mature broad positive record. A clean record that is still thin stays Insufficient Record.
• Insufficient Record: no scored evidence, no votes, unresolved vote context, fewer votes than the office minimum, or a clean thin record. The office minimum is the lower of 100 and half a typical full record for that office, and it is never below 8. A typical full record is the loaded-vote count at the 75th percentile of members of that office who have at least one vote.
• No Positive Choice at the race level when no candidate clears the standard

A Low confidence (agenda resampling) note means the Unfit, Compromised, or Qualified label agrees in fewer than 90 percent of 200 resamples of that chamber's scored bills. Disqualified and Insufficient Record do not receive this check.

A Borderline badge marks an Unfit, Compromised or Qualified label whose Tier 2 scale score is within 3 points of the Compromised (25) or Unfit (50) cutoff, so the label could change with small differences in the record; the label itself is unchanged.

• It uses deterministic scoring rather than black-box rankings.
• It keeps the moral rules separate from the LLM tagging layer.
• It weighs different kinds of votes differently instead of treating every vote as equally revealing.
• It handles multiple votes on the same bill without unfair double-counting.
• It normalizes similar bill tags into a smaller set of reviewed scoring categories.
• It ranks candidates from roll-call votes and from reviewed sponsorship and cosponsorship, not from campaign money, lobbying, or affiliations.
• It is designed to show evidence for why a candidate received a status instead of outputting a mysterious score.

For a fuller explanation of how votes become scores, see the scoring mechanics. For the biblical rationale and Scripture references behind those judgments, see the Doctrine page.

Not covered yet

• Challengers and newcomers are not named and are not scored.
• Presidents, governors, and other executives are not graded from this legislative record. Presidential filings can appear in the office counts. In this snapshot those rows are Insufficient Record.
• State legislators are not covered.
• Committee votes are not covered.
• Presidential nominations are not graded.

Fairness

A positive-fitness credit is withheld when at least 70 percent of the member's caucus cast the same vote, in a cohort of at least 10 (Proverbs 24:11). The rule does not withhold penalties and it never sets a label. Life, child protection, and the other Tier 1-grade concerns stay exempt. An independent is counted with the caucus that person votes with at least 75 percent of the time, over at least 20 votes.

• Procedural votes can be difficult to interpret when the bill context is thin or the vote intent is unclear.
• Drafted tags can be imperfect. A bill marked needs review does not score until a person reviews it.
• A candidate's result may look stronger or weaker depending on how much verified legislative data is actually available for that person.

JustMeasure is meant to improve clarity, not pretend to offer perfect knowledge. Where the record is thin, the system is designed to surface that limitation rather than hide it.