Skip to main content

ValiFit Constitutional Charter

Algorithmic Transparency

The principle

Every scoring formula is public, every weight is documented, every methodology is reproducible from federal data.

The commitment

The 8 ValiFit composite scores (Budget, Commute, Appreciation, School Investment, Environmental Safety, Safety Infrastructure, Health Infrastructure, Livability) each have public methodology pages at /trust/methodology and federal-source documentation at /trust/sources. The unified-scorer source is in lib/composites/* — a single function per composite, no hidden ML model, no black-box ranking. Anyone can trace a 78/100 score back to its component inputs (each cited) and verify the math. This is opposite of Zillow's Zestimate, which is a regression model trained on private MLS feeds with weights nobody can inspect.

Why this matters

Black-box scoring lets companies hide bias. If you can't see the weights, you can't audit the bias. We make the weights public so the bias audit is mechanical: anyone can run the formula on any neighborhood and get the same result we did.

How we enforce

lib/composites/* — one TS file per composite, no compiled binaries, no encrypted weights, full unit-test coverage. /trust/methodology renders the formulas. Every composite output ships with explanation: { weights, inputs, source } so the math is reconstructible.

Aligns with

  • NIST AI Risk Management Framework
  • EU AI Act Article 13 (transparency)

Last reviewed and affirmed: 2026-05-08. The charter is versioned with the codebase — a Git history of changes to lib/charters.ts serves as the audit log.

Other charters