Skip to content
Keystone Check

Transparency

How these grades are built

Every claim on a member's profile traces back to a record we can point you to. This page documents the whole procedure, v1.4.0: what we count, how it is weighted, where it comes from, and what it cannot tell you.

What counts as evidence

Only a legislator's own official acts and words count. No records have been ingested yet, across any of the 252 seated members. Endorsements, party affiliation, and third-party commentary about a member are deliberately excluded, because they describe how others see a legislator rather than what the legislator did.

    How records are weighted

    Not every record says the same amount about a member. A recorded floor vote is near-conclusive; an offhand remark in a hearing is suggestive at best. Each record therefore carries a weight between 0 and 1. No records have been ingested yet, so there is no observed range to report.

    Roll-call votes and sponsorships carry the heaviest weights. They are recorded, deliberate, and hard to walk back.

    Hearing remarks and cosponsorships sit in the middle. They are meaningful, but weaker signals of commitment.

    Press coverage is weighted lowest. It is the only category that records what someone else said a legislator meant, so it can suggest a direction but never settle one.

    The grade scale

    A subject score runs from 0 to 100 and maps to a letter on the standard scale below. These are the exact thresholds the code uses.

    97+
    93+
    90+
    87+
    83+
    80+
    77+
    73+
    70+
    67+
    63+
    60+
    0+

    A member with no qualifying record on a subject is marked NR, meaning not rated. That is a gap in our coverage, not a judgment about them, and it is never counted as a failing grade.

    Alignment records are published; letter grades are not

    8 roll calls from the 2023-2024 session are scored against the model below, producing records for 225 of 258 seated members. 34 members cast no recorded position on any scored vote and read NR, a gap in the legislative record, not a judgment about them.

    Letter grades are withheld for the 2023-2024 session. It produced at most four scored votes per member, too few distinct outcomes to justify thirteen letter bands. Per-vote alignment records are published instead.

    The scoring model

    These positions are ours, not the legislature's

    The votes below are facts from the public record. Deciding that a yes on a given bill is the favorable direction is an editorial judgment made by Keystone Check. No legislator, agency, or advocacy organization holds these positions on our behalf, and a member scored as opposed may simply disagree with us.

    Model 1.0.0, adopted 2026-08-04. Grades rest on 8 hand-selected roll calls from the 2023-2024 session. There is no keyword classifier: every scored vote is written out by hand, because a wrong direction silently inverts a real person's record.

    Why there are no letter grades this session

    Letter grades are withheld for the 2023-2024 session. It produced at most four scored votes per member, too few distinct outcomes to justify thirteen letter bands. Per-vote alignment records are published instead.

    Across 225 members with evidence, the model can only produce 5 distinct scores (50, 67, 74, 75, 100). Thirteen letter bands cannot rest on that, so each member gets an alignment record instead: which bills they voted on, how they voted, and whether that matched our position.

    Scored votes
    HB 1476Mining & Energyfavorable: no

    Our position: Crypto-asset mining operations should not be singled out for reporting obligations and impact studies that comparable industrial energy users do not carry.

    HB 1476 imposes reporting requirements and an impact study specifically on qualifying crypto-asset mining operations. Scored as a targeted compliance burden on the sector, so a NO vote is treated as aligned. It passed the House 102-101, a genuine one-vote margin, and the most contested digital-asset vote of the session.

    HB 2481Digital Assetsfavorable: yes

    Our position: Pennsylvanians should retain the right to self-custody digital assets and to operate blockchain nodes without state restriction.

    HB 2481 restricts government interference with the use and custody of digital assets and expressly authorizes node operation. A YES vote is self-custody protective. Passed the House 176-26.

    HB 2081Cryptocurrencyfavorable: yes

    Our position: Virtual-currency lenders holding customer assets should be required to segregate customer funds and meet collateral requirements.

    HB 2081 imposes collateral requirements and fund segregation on virtual currency lenders. Scored as consumer protection against commingling of customer assets, so a YES vote is aligned. Committee-only, so weighted at 0.35.

    SB 1320FinTechfavorable: yes

    Our position: Money transmission licensing should be modernized to give payment and fintech businesses a workable, current licensing path.

    SB 1320 amends the Money Transmission Business Licensing Law. A YES vote is treated as supporting a modernized licensing regime. Passed the Senate 50-0; unanimity means it separates almost no one, and it is included for completeness of the record rather than for discrimination between members.

    SB 633FinTechfavorable: yes

    Our position: A regulatory sandbox should let financial and technology firms pilot products under supervision rather than face the full licensing burden up front.

    SB 633 establishes the Regulatory Sandbox Program and Regulatory Relief Office. A YES vote supports supervised experimentation. Committee-only, weighted 0.35.

    HB 1598Artificial Intelligencefavorable: yes

    Our position: Consumers should be protected from deceptive uses of synthetic and AI-generated media under the Unfair Trade Practices and Consumer Protection Law.

    HB 1598 extends consumer-protection law to cover AI-generated deceptive content. A YES vote is treated as pro-consumer-disclosure. Passed the House 146-54.

    HR 170Artificial Intelligencefavorable: yes

    Our position: The Commonwealth should study AI’s effects on labor, privacy, and public services before legislating broadly on it.

    HR 170 directs the Joint State Government Commission to convene an advisory committee to study artificial intelligence. A YES vote supports evidence gathering ahead of regulation. Adopted 125-77. Only the final adoption vote is scored; the three amendment votes on the same day are procedural and excluded.

    SB 563Cybersecurityfavorable: yes

    Our position: Deploying ransomware against Pennsylvania systems should be a distinct criminal offense.

    SB 563 creates a ransomware offense in Title 18. A YES vote supports explicit criminal liability for ransomware operators. Committee-only, weighted 0.35.

    Weighting

    A floor vote weighs 1.00 and a committee vote 0.35. Floor votes put the whole chamber on the record; committee votes involve 11 to 25 members and are often procedural, so they inform a grade without carrying it.

    A topic needs at least 1.00 weighted points before it is scored at all. Below that it is reported as having no record rather than being scored, because one committee vote is an anecdote, not a position.

    What this model cannot tell you

    Topics with no votes on record: Stablecoins, CBDCs, Blockchain. The General Assembly did not hold a recorded vote on these in 2023-2024. They are shown as absent, never as zero: a member with no stablecoin votes has no stablecoin record, which is not the same as a bad one.

    Unattributable votes.About 12% of individual positions in the upstream data carry no voter identifier. They count toward a bill's official tally but toward no member's grade, because matching them by name would risk assigning a vote to the wrong legislator, four different members of this session share the surname Williams.

    On-topic bills we deliberately did not score

    HB 1139: Amends the Administrative Code in the organization of departmental boards. Reached only via a cybersecurity keyword match on a broad administrative bill; the operative subject is agency structure, not cyber policy, so scoring it would attribute a cyber position to members who were voting on something else.

    HB 1282: Computer Data Center Equipment Incentive Program tax provisions. Data centers are adjacent to crypto mining but not equivalent, and a vote on a broad data center tax incentive does not establish a position on mining.

    SB 654: A Tax Reform Code sales-and-use-tax exclusions bill that touches data center equipment among many unrelated items. With 16 roll calls spanning budget negotiations, a vote here reflects tax and budget politics rather than any digital-asset position.

    Verified source data
    Public record
    Recorded as published by the source. Keystone Check does not alter these values.
    County geography
    US Census Bureau cartographic boundary files, all 67 counties.
    Legislator roster
    Chamber, district, and seat assignments for every profiled member.
    Bills and actions
    Bill text, status, and procedural history.
    Roll-call votes
    Recorded floor and committee votes as published.
    Official statements
    Attributed public remarks with a citable source.
    Keystone Check analysis
    Our judgement
    Derived by Keystone Check from the data on the left. Not issued, endorsed, or reviewed by the Commonwealth of Pennsylvania.
    Policy classification
    Which bills and statements count as digital-asset evidence.
    Evidence weighting
    How a vote is weighed against a sponsorship or a statement.
    Letter grades
    The composite score and its A–F letter.
    AI summaries
    Machine-written précis, labelled wherever shown.
    Read how grades are calculated

    Data sources

    Records are gathered from the Pennsylvania General Assembly and contemporaneous reporting. Every record on a member’s profile is required to show where it came from, though no records have been ingested yet.

    Corrections

    Grades are only as good as the records under them, and records can be wrong: a misattributed vote, a quote stripped of context, a bill that changed after we read it.

    How review works

    Every record carries a review status of approved, pending, or needs review. That status is shown on the record itself wherever it appears, so you can see which claims we have checked against the source and which we have not.

    The intended production behaviour is that only approved records feed a score, so correcting a record updates the grade with no editorial override. Note that this preview does not yet enforce that: review status is currently displayed but does not gate scoring.

    Found something wrong? Every record links to its source so you can check our work directly. Start from a member's profile or ask the archive what a grade rests on.