How The 42 Engine Works
Full transparency on how we score legislation. No black boxes. Every number is explained.
๐ณ๏ธ How It Works โ Plain English
Step 1: AI Reads the Bill (Specific Version)
When a new bill is introduced in Congress, our system automatically downloads the full text. An AI analyst (powered by Google Gemini) reads the bill and scores it across 5 core pillars โ the same 5 dimensions every piece of legislation affects, whether it's a tax cut or a climate treaty. Because bills change rapidly through amendments, the system always explicitly lists the specific version of the bill being scored (e.g., "Engrossed in House," "Senate Amendment"). The AI doesn't score based on popular opinion or political party โ it scores based purely on what the bill's legal text actually does.
Step 2: You Set Your Values (With a 90-Day Lock)
In your Settings, you allocate 100 points across the 5 pillars based on what matters most to you. If you care deeply about Personal Liberty but less about National Security, your allocation reflects that. This is your Profile Vector โ your mathematical DNA as a voter. To prevent reactionary gaming of the system during a highly publicized vote, users can only update these core weights once every 90 days, enforcing long-term ideological consistency.
Step 3: Your Vote Gets Weighted
When you vote YAY or NAY, the system calculates how much this specific bill matters to you personally. If a bill scores +80 on Economic Growth and you allocated 40 points to Economic Growth, your vote carries more weight on that bill than someone who only allocated 5 points. This is your Salience Score โ it ensures that the legislation is weighted by its salience relative to your stated priorities.
Step 4: The Consensus Score
All weighted votes are summed together and normalized to a -100 to +100 scale to produce the Consensus Score. Positive means the collective leans YAY, negative means NAY. The score is divided by the total weight pool so it stays interpretable regardless of how many people have voted: +100 = everyone unanimously supports with full value alignment, -100 = everyone unanimously opposes. The total number of votes is always displayed alongside the score for full transparency.
โ Frequently Asked Questions
Can the AI be biased?
Any AI can have biases. That's why we publish the exact scoring rubric (see below), require the AI to justify every score by citing specific criteria, and display those justifications on every bill. If a score seems wrong, you can see exactly why it was given.
Why does my vote weight vary from bill to bill?
Because different bills affect different pillars. A defense spending bill triggers your National Security weight, while a privacy bill triggers your Personal Liberty weight. Your total influence is proportional to how much you care about the specific issues a bill touches.
Can politicians see this data?
Yes โ that's the point. The consensus score is designed to be a transparent, mathematical signal that representatives can reference to understand where their constituents actually stand on legislation, weighted by personal impact.
Why can I only change my Profile Vector every 90 days?
This "Profile Vector Lock" is an anti-gaming mechanism. It prevents users from suddenly shifting all their weights to 100 on a single pillar just to maximize their voting power on a specific, highly publicized bill. It ensures the platform measures your true, long-term ideological baseline.
๐ Technical Documentation
Scoring Rules
These rules are injected directly into the AI's system prompt. Every score must comply with them.
- The Doctrine of Uncertainty: Score ONLY what the statutory text explicitly does. If a billโs language is deliberately vague or delegates sweeping authority to an agency without clear guardrails, you MUST state this uncertainty in your justification and score based on the highest-probability enacted outcome.
- Score relative to the STATUS QUO. A bill that maintains current law scores 0.
- Each pillar is scored INDEPENDENTLY. A bill can score +80 on Economic and -60 on Sustainability.
- Use the band criteria as your anchor. First identify which band the bill falls into, THEN select a specific number within that band.
- Symbolic resolutions with no legal mechanism score 0 on all pillars they do not affect.
- You MUST provide a 1-sentence justification per pillar citing the specific band criteria that determined your score.
- Err toward the middle of a band rather than the extremes unless the evidence is overwhelming.
- If a bill has multiple provisions affecting the same pillar in opposite directions, net them together and explain the trade-off in your justification.
5-Pillar Scoring Rubric
The V1 Normative Framework
We recognize that these 5 pillars are a chosen normative framework, not a permanent claim to universal truth. They reflect the V1 consensus on the foundational tensions of American governance. As the cooperative matures, any verified user can propose adding, removing, or restructuring a pillar via the Pillar Amendment Process (requiring supermajority ratification).
Each bill is scored from -100 to +100 on five independent pillars. The AI must identify which score band the bill falls into based on the criteria below, then select a specific number within that band.
+Growth & Opportunity โ โ -Burden & Restriction
Scoring question: "Does this bill make it easier or harder for businesses and workers to generate wealth?"
| Score | Criteria |
|---|---|
| +80 to +100 | Eliminates a major tax or regulatory barrier; opens entirely new markets or industries.e.g., Repealing a nationwide industry tax. |
| +40 to +79 | Significant tax cuts, deregulation, trade expansion, or direct job creation programs.e.g., Small Business Lending Act providing tax credits to startups. |
| +1 to +39 | Minor economic incentives: targeted tax breaks, grants, or modest deregulation.e.g., Tax credit for rural broadband investment. |
| 0 | No measurable economic impact. Symbolic resolutions, ceremonial designations.e.g., Declaring National Ice Cream Day. |
| -1 to -39 | Minor compliance costs, new reporting requirements, or modest fee increases.e.g., New labeling requirements for food products. |
| -40 to -79 | Substantial new taxes, industry-specific regulations, or mandated spending.e.g., Cap-and-trade carbon pricing system. |
| -80 to -100 | Bans an entire industry, imposes sweeping price controls, or causes structural economic disruption.e.g., Outright ban on fossil fuel extraction. |
Calibration Examples
+Freedom & Privacy โ โ -Restriction & Mandate
Scoring question: "Does this bill give individuals more control over their own lives, or less?"
| Score | Criteria |
|---|---|
| +80 to +100 | Repeals a law that criminalized individual behavior; eliminates mass surveillance programs.e.g., Repealing the Patriot Act's mass data collection provisions. |
| +40 to +79 | Expands individual rights, adds opt-out provisions, strengthens due process protections.e.g., Right-to-repair legislation letting consumers fix their own devices. |
| +1 to +39 | Minor privacy protections, transparency requirements, or speech protections.e.g., Requiring warrants for email search by law enforcement. |
| 0 | No impact on individual liberty or personal choice.e.g., Renaming a post office. |
| -1 to -39 | Minor compliance mandates, registration requirements, or opt-in defaults.e.g., Mandatory voter ID with free ID provision. |
| -40 to -79 | Mandates behavior, restricts personal choice, requires government approval for individual activity, or endorses one group's beliefs through government.e.g., Government-mandated display of religious text in public buildings. |
| -80 to -100 | Criminalizes previously legal individual behavior, eliminates a constitutional protection, or enables warrantless surveillance.e.g., A bill criminalizing end-to-end encryption. |
Calibration Examples
+Fairness & Inclusion โ โ -Exclusion & Inequality
Scoring question: "Does this bill make society more fair for historically disadvantaged groups, or less?"
| Score | Criteria |
|---|---|
| +80 to +100 | Establishes new constitutional-level protections for a marginalized group; closes a major systemic inequality gap.e.g., Passing a constitutional amendment guaranteeing equal rights regardless of gender. |
| +40 to +79 | Expands civil rights protections, mandates equal pay, or funds programs targeting underserved communities.e.g., The Equality Act expanding anti-discrimination protections to LGBTQ+ individuals. |
| +1 to +39 | Minor equity improvements: diversity reporting, accessibility standards, or cultural recognition.e.g., Funding for minority-language ballot access. |
| 0 | No equity impact. Affects all demographics equally or is purely procedural.e.g., Adjusting the federal fiscal year calendar. |
| -1 to -39 | Inadvertently widens an existing gap or removes a minor protection.e.g., Eliminating a small grant program for rural schools. |
| -40 to -79 | Actively removes civil rights protections, defunds equity programs, or restricts voting access.e.g., Removing Section 5 preclearance from the Voting Rights Act. |
| -80 to -100 | Legalizes discrimination, creates separate legal standards for a specific group, or eliminates foundational civil rights.e.g., A bill legalizing racial discrimination in hiring. |
Calibration Examples
+Safety & Defense โ โ -Vulnerability & Risk
Scoring question: "Does this bill make the country safer or more vulnerable?"
| Score | Criteria |
|---|---|
| +80 to +100 | Major military modernization, alliance treaty, or critical infrastructure defense program.e.g., NATO mutual defense reaffirmation treaty. |
| +40 to +79 | Significant defense funding, cybersecurity programs, or intelligence capability expansion.e.g., Cybersecurity Infrastructure and Security Agency funding increase. |
| +1 to +39 | Minor security improvements: equipment upgrades, training programs, or diplomatic initiatives.e.g., Funding for local police body camera programs. |
| 0 | No national security impact.e.g., A resolution about public art. |
| -1 to -39 | Minor security trade-offs: transparency requirements that could reveal capabilities.e.g., Requiring public disclosure of drone strike statistics. |
| -40 to -79 | Significant military drawdowns, intelligence restrictions, or weakening of border controls.e.g., Cutting defense budget by 25% without strategic transition plan. |
| -80 to -100 | Dissolves a military alliance, eliminates a critical intelligence capability, or compromises nuclear deterrence.e.g., Unilateral withdrawal from NATO. |
Calibration Examples
+Long-term Investment โ โ -Short-termism & Extraction
Scoring question: "Does this bill leave the country better or worse off in 30 years?"
| Score | Criteria |
|---|---|
| +80 to +100 | Transformational long-term investment: nationwide renewable energy transition, major education overhaul.e.g., A Green New Deal-scale climate investment bill. |
| +40 to +79 | Significant climate, infrastructure, or R&D investment with measurable long-term returns.e.g., Federal funding for nationwide EV charging infrastructure. |
| +1 to +39 | Minor environmental protections, small R&D grants, or incremental infrastructure improvements.e.g., Tax credit for residential solar panels. |
| 0 | No long-term sustainability impact.e.g., A resolution commemorating Earth Day. |
| -1 to -39 | Minor rollback of environmental protections or short-term fiscal prioritization.e.g., Exempting small businesses from emissions reporting. |
| -40 to -79 | Repeals major environmental regulations, expands fossil fuel extraction, or defunds climate research.e.g., Repealing the Clean Power Plan. |
| -80 to -100 | Eliminates foundational environmental law, opens protected lands to unrestricted extraction.e.g., Repealing the Endangered Species Act and opening all national parks to drilling. |
Calibration Examples
Consensus Formula
Sแตข = ฮฃ (UserPillarWeight_k ร BillPillarScore_k) for k = 1 to 5The dot product of a user's personal value allocation and the bill's pillar scores. The absolute value |Sแตข| is used as the vote's magnitude โ it represents how much this bill matters to this person, regardless of whether the bill helps or hurts their priorities.
Consensus = ฮฃ (Vแตข ร |Sแตข|) / ฮฃ |Sแตข| ร 100Each vote's direction (Vแตข = +1 for YAY, -1 for NAY) is multiplied by its magnitude |Sแตข|, summed, then divided by the total weight pool and scaled to -100 to +100. This normalization ensures the score is always interpretable regardless of how many people have voted: +100 = unanimous support with full alignment, -100 = unanimous opposition with full alignment.
AI Model & Pipeline
Model
Google Gemini Flash (latest)
Output Format
Structured JSON via Zod schema validation
Data Source
Congress.gov API (official bill text)
Scoring Method
Band-anchored rubric with calibration examples
Transparency
Per-pillar justifications stored and displayed
Refresh Rate
Every 24 hours (automated background worker)
Known Limitation: Single-Model Evaluation
The V1 platform relies on a single capable model (Gemini Flash) to maintain sustainable API costs. We recognize that any single LLM contains implicit biases. Upgrading to a multi-model "Ensemble Pipeline" (aggregating scores across DeepSeek, Claude, and OpenAI) is a critical priority on our 12-month roadmap.
โ๏ธ The Dissent & Nuance Engine (Planned)
The Problem: Majority Blind Spots
A raw consensus score captures the will of the majority, but it often flattens nuance and ignores valid critiques from the minority. A healthy democratic system must actively elevate dissenting perspectives, rather than burying them under downvotes.
Our Two-Part Solution
1. Dissenting Perspective Panels
Upon the conclusion of a legislative poll, the system does not just display the winning consensus. It actively curates and pins a "Dissenting Panel" โ the top three most highly rated, fact-checked arguments from users who voted on the losing side. We force the winning majority to look the best opposing arguments in the eye.
2. Minority Impact Reports
For major legislation, the AI generates a post-vote analysis focusing explicitly on the user cohorts with the lowest Salience Scores. How does this bill affect the people whose profile weights were least aligned with its impacts? This surfaces blind spots and unintended consequences for the users who were mathematically marginalized on a specific vote.
Why Not Now?
Building these analytical tools requires a robust dataset of user votes and discourse. We believe in getting the foundation right first โ transparent scoring, value-weighted voting, and verifiable identity โ before layering on automated dissent analysis.
This methodology document is a living document. The scoring rubric, formula, and AI model are subject to refinement as we learn from community feedback.
Last updated: March 2026