How We Score
v2.1Every weight, threshold, and calculation is published here. No black box. The numbers below are pulled live from the algorithm configuration — the same values used to compute every score on the platform.
Light of Liberty only scores what it can read in full. No clips. No excerpts. No secondhand summaries. If the platform doesn't have the complete source material, it doesn't score it. Every score links to the full, verified source document so you can read it yourself.
Version 1 produced a single score (0-100) per domain. Version 2 replaces this with Fidelity Vectors — three-dimensional measurements at the level of individual constitutional provisions.
Fidelity Vectors
The heart of v2. Every piece of source material is scored against each relevant constitutional provision using a five-dimensional vector. The first three dimensions capture fundamentally different aspects of constitutional fidelity:
Does this action reinforce or contradict the provision as written? +1.0 = directly reinforces, -1.0 = directly contradicts
How far does the reasoning travel from the actual text? 0.0 = directly supported, 1.0 = no textual support (penumbras, emanations)
Does the action respect separation of powers and federalism? +1.0 = perfect structural respect, -1.0 = usurps another branch or level
Plus two weighting dimensions that determine how much each provision contributes to the composite score:
How certain is this analysis? A vote on a bill that explicitly regulates firearms is straightforward to measure against the Second Amendment — confidence approaches 1.0. A regulation about digital data retention implicating the Fourth Amendment's “papers and effects” is genuinely ambiguous — confidence might be 0.6. Lower-confidence scores carry less weight in the composite so that uncertain assessments don't distort a figure's overall rating. The citizen always sees the confidence level — we never pretend to have certainty we don't have.
How directly does this action touch this provision? A bill that explicitly regulates speech has high First Amendment salience (near 1.0). A bill that tangentially affects speech through a complex regulatory chain has low salience (perhaps 0.2). Crucially, salience measures the connection between the action and the provision — not the importance of the provision itself. The algorithm never decides which rights matter more. It only measures how directly each action engages each right. No thumb on the scale.
The Three Axes: Say · Do · Fund
Every public figure is measured on three dimensions. What they say matters — but what they actually do matters more. And who funds them reveals potential influences on both.
Speeches, campaign promises, official statements, press releases, social media (complete threads only)
Votes, legislation sponsored/cosponsored, executive orders signed, absences from constitutionally significant votes
Campaign contributions, PAC affiliations, lobbying disclosures — who pays and what changes after
Signal strength within DO: Sponsoring > Cosponsoring > Voting > Absence (scored as dereliction when constitutionally significant)
Composite Score
For display purposes, fidelity vectors are converted to a 0-100 composite. The three dimensions contribute equally, weighted by salience and confidence per provision.
The Say-Do Gap
The most telling metric: the gap between what a figure says and what they actually do. A politician who talks about constitutional principles but votes against them will have this exposed clearly, with links to every piece of source material.
Semantic Drift Detection
Certain words in the founding documents had specific meanings at ratification that differ from modern usage. The Constitutional Lexicon tracks these terms and flags when source material uses them in their drifted modern sense rather than the original meaning. This catches subtle constitutional erosion that alignment scoring alone would miss.
Deviation Index
Beyond scoring individual actions, the engine tracks how constitutional meaning has been stretched, contracted, or reinterpreted over time. Each deviation is recorded, typed, and linked into chains so citizens can trace how one departure enabled the next.
Each deviation is scored 0.0–1.0 for magnitude, linked to affected provisions, and chained to prior deviations. Corrections are tracked separately — when a deviation is walked back, the record shows it.
Advanced Scoring Dimensions
Beyond vote-by-vote scoring, the engine runs additional analytical passes that capture patterns invisible in any single action.
The Oath: What They Swore
Every federal official takes an oath to the Constitution. The oath is not symbolic — it is a binding commitment and the starting point of every fidelity timeline. Different offices use different statutory oath texts, but all share the same core obligation: defend the Constitution.
“I do solemnly swear (or affirm) that I will support and defend the Constitution of the United States against all enemies, foreign and domestic; that I will bear true faith and allegiance to the same; that I take this obligation freely, without any mental reservation or purpose of evasion; and that I will well and faithfully discharge the duties of the office on which I am about to enter: So help me God.”Source: 5 U.S.C. § 3331 (based on Article VI, Cl. 3)
“I do solemnly swear (or affirm) that I will faithfully execute the Office of President of the United States, and will to the best of my Ability, preserve, protect and defend the Constitution of the United States.”Source: Article II, Section 1, Clause 8 (verbatim from Constitution)
“I do solemnly swear (or affirm) that I will administer justice without respect to persons, and do equal right to the poor and to the rich, and that I will faithfully and impartially discharge and perform all the duties incumbent upon me under the Constitution and laws of the United States: So help me God.”Source: 28 U.S.C. § 453
Trajectory: Person vs. Themselves
Figures are never compared to each other. They're compared to themselves over time, measured against the fixed standard of the founding documents. The trajectory shows whether they're moving toward or away from constitutional alignment — and at what speed.
Anti-Bias Architecture
Six-Layer Data Pipeline
Source material flows through six distinct layers — each with its own validation — before a score ever reaches a citizen's screen.
Live Algorithm Configuration
These values are loaded directly from the production database — the exact same numbers that compute every score.
Loading configuration…
Data Sources
Scores are built from official, verifiable sources. Each score event links directly to the complete source material — read it yourself.