Built for oversight · Designed for trust

AI, Transparency & Trust

CivicPrinciples uses artificial intelligence to support civic transparency and accountability. The platform is developed and operated by Civix IQ. (Integrity Matters is an independent civic-engagement organization that may partner with or promote the platform as part of its public-accountability mission.) The platform evaluates local government activity against Civix IQ's 15 core principles of good governance.

This technology does not replace human judgment — it complements it. By analyzing official records, meeting transcripts, and votes, CivicPrinciples offers a clear, accessible view of how your local government operates.

Who wrote the standard, and who applies it

CivicPrinciples is the product of two distinct kinds of work, and we think it matters to keep them separate.

The methodology is human. The 15 principles of good governance, the severity classifications (Clear / Likely / Potential), the type weights, and the editorial standards that govern what we publish were all designed by people, in consultation with governance experts and advisors. The methodology reflects deliberate, considered choices about what good local governance looks like.

The application is AI. Once the framework is set, Civix IQ's analysis engine applies it to every action consistently — reading the public record, classifying votes against the principles, and generating the assessments you see published.

This separation is intentional. Every member is evaluated by the same standard, in the same way, every time. The scores and write-ups are generated through the consistent application of the published governance framework rather than ad hoc personal or political judgment. We stand behind that framework, and we welcome scrutiny of it.

How the AI sees what matters

Our system scans and organizes public records from meetings: votes, motions, agendas, and spoken transcripts. It then applies a structured evaluation model to assess how each action aligns with principles such as transparency, fairness, and fiscal responsibility. At the heart of this model is a combination of:

  • Natural language processing (NLP) techniques trained on public meeting data
  • Rules and scoring logic derived from Civix IQ's governance framework
  • Continuous human oversight and calibration

The result: a focused, principle-based perspective. Not a data dump. Not a black box.

How the Screw-O-Meter works

CivicPrinciples uses a structured rubric to evaluate every action, reflecting both the frequency and severity of potential governance violations. The goal is not to sensationalize, but to help the public see patterns that matter. We assess each action using three levels of concern:

  • Clearly Violates — the action demonstrably conflicts with one or more core principles, with clear evidence of harm.
  • Likely Violates — there is strong evidence the action conflicts with a principle, though context may vary.
  • Potentially Violates — the action may conflict with a principle, but the evidence is limited or unclear.

Every vote has two sources of potential violation, and our scoring reflects both:

  • What the ordinance does — the substance of the policy itself: who it benefits, whose rights it affects, what burdens it imposes.
  • How the body handled it — the conduct around the vote: what testimony was heard, which concerns were engaged with or dismissed, what members said on the record, what alternatives were considered.

CivicPrinciples evaluates each vote based on the full record before the body, including both the substance of the policy and contextual factors such as conduct, public discussion, testimony, and the overall decision-making process.

Each level of concern carries a weight, and each evaluated member accumulates points based on their voting record. The system separately considers how often a member's actions raise concerns and how severe those concerns are. These are combined into a final score between 0 and 5 — where 0 represents full alignment with governance principles and 5 reflects the most serious and frequent violations.

We call this score the Screw-O-Meter: a visual, data-backed indicator of how consistently a member's decisions align with our standard for good governance. While the scoring methodology is proprietary, it is consistently applied, evidence-based, and applied equally to all members. We disclose the criteria and classification framework that drive the score, and we support confidential third-party methodology reviews by qualified neutral experts when appropriate.

Two ways we evaluate government

CivicPrinciples evaluates government at both the individual and institutional level.

Member Ratings measure how individual members vote across contested issues and how consistently those decisions align with our governance principles.

Body Performance evaluates the governance quality of the body's collective decisions by analyzing whether majority-supported decisions conflict with governance principles.

Although the two systems draw from the same principle assessments, they answer different questions and can move independently of one another.

Why member ratings and body performance can differ

A body can contain several poorly-rated members and still receive a Principled rating if those members are consistently outvoted by the majority.

Likewise, a body can contain several well-rated members and still receive a Screwed rating if majority voting patterns repeatedly produce decisions that conflict with governance principles.

The two systems are designed to separately evaluate individual voting behavior and collective governance decisions.

Transparency about limitations

We believe it's just as important to acknowledge what AI can't do. The system may occasionally misclassify statements, miss nuances in tone or sarcasm, or display minor transcription errors. We do not treat its output as infallible — instead we treat it as a high-quality input for public discussion and civic engagement.

When errors occur, they are not ignored. We welcome feedback and corrections through Report an Issue.

Bias, safeguards, and oversight

Civix IQ's models are designed to apply the governance framework consistently across officials, parties, and policy positions. The system is not designed to favor political parties or ideological outcomes. The principles themselves were developed in collaboration with governance experts and nonprofit advisors to reflect widely recognized governance standards. In addition:

  • Every new model iteration is reviewed for potential bias
  • Flagged content is routinely audited by humans
  • All scoring criteria are documented and published

Learn more about the 15 Principles.

A living system for civic engagement

Technology changes. Governing bodies evolve. That's why CivicPrinciples is built as a living system — designed to improve over time, adapt to local needs, and serve the public good with humility and clarity.

If you ever have questions about how the system works or where it may fall short, we invite you to ask. Report a concern or contact us — we take every message seriously.

CivicPrinciples is powered by Civix IQ, which manages the platform's analysis engine and evaluation framework. Civix IQ welcomes meaningful review. When members or institutional partners raise good-faith questions about scoring fairness, we support third-party audits conducted under NDA by neutral experts. Audits review the consistent application of our evaluation framework — not political positions — and help us uphold fairness, accuracy, and public trust. Learn more in our Audit Policy, and see our Disclaimer for the limits of AI-generated content.