Glossary

A dimension of ideology, asked as a question. There are seven: Who owns, Who provides, Who pays for harm, Who decides, Who counts, Who has a say, and Who sets the terms. Each is scored independently from −4 to +4, with 0 at its centre. A question the article does not take up is left off its card and named once on its analysis page. An article can be left on one axis and right on another. The scores are never collapsed or averaged. Everyday fights such as guns, immigration, housing and wages are topics, not axes, and each is read on whichever questions it takes up.

The full theoretical range on an axis, from one philosophical endpoint to the other. Defined by political philosophy, not by the current media landscape. On Who owns, it runs from “Nobody owns a business” at −4 to “Everything privately owned” at +4. The spectrum is permanent — it does not shift. The Overton window moves within it.

The scoring scale that defines what each integer level means on each axis. Each of the nine rungs on a question has a name, a description of the arrangement it stands for, and a guide to what an article sounds like there. "+1 on Who owns" is the rung called “Private ownership, exclusive and unlimited.”

A position on an axis, given as a whole number from −4 to +4. Where the three readings of an article are further apart than close agreement allows, the axis is shown as a range instead of one number. A score alone is incomplete — it needs mode and engagement to be meaningful.

Whether the article argued the position (explicit), assumed it without stating it (implicit), or did both (mixed). Implicit positions are often more diagnostic because they reveal what the writer treats as the unmarked default.

How centrally the article engages a given axis, scored 1 to 4. An article can score +2 on Who owns with engagement 1 (it barely touches the question) or engagement 4 (the entire article is about who owns what). Both are +2. They are not the same article.

An observational instrument that detects a specific kind of signal in text. There are 14 lenses organized in 6 depth tiers, from surface-level framing down to narrative continuity. Each lens uses a "Category: Parameter / Parameter" naming format that describes what it measures.

A grouping of lenses by how deep they look into the text. Six tiers: Surface, Frame, Structure, Context, Absence, Continuity. The numbering of the lenses follows the tiers — Lens 1–2 are Surface, Lens 3–5 are Frame, and so on. The numbering IS the process.

What a lens produces when it examines an article. A specific, concrete finding tagged to one or more axes. "The article centers the founder's vision without mentioning the workforce" is an observation. "+2 on Who owns" is NOT an observation — that is a score, which comes later.

The passage of the article used as evidence for a score, described in our own words. Where the exact wording is itself the evidence, an exhibit quotes it: exactly, in no more than 15 words, and never more than five times in one analysis. Each exhibit is tagged with the lens that identified it and the axis it supports.

The tool's judgment on one axis: which arrangement the article treats as proper, and which rung of the rubric that arrangement fits. Every article is read three times, independently. When the readings agree closely on an axis (two on the same rung, and the third on it, one step from it, or at No Signal), that rung is the score. When they are further apart and at least two found a position, the axis is shown as a range, and the page explains what each reading found.

The full audit trail connecting a score back to the article text: Score → Rubric level → Exhibit → Observation → Lens → Article text. If someone disagrees with a score, the evidence chain tells them exactly where their disagreement sits.

A vetted reference point in the scoring library. A real-world figure, law, institution, or court decision whose position on the axes has been determined through the full evidence-first process and has passed the vetting chain (internal review → scholar review → co-publication → versioned release). The anchors were scored on the five axes the rubric used until October 2026, and they are out of use until they are re-scored on the seven questions. Until then, articles are not measured against them.

The complete set of vetted anchors, tracked by policy domain. It was mapped across the earlier five axes and nine score levels. It is out of the menu until it is re-scored on the seven questions.

A visualization showing where anchors exist across all axes and score levels, and where gaps remain. Gaps trigger the "scavenger hunt" — targeted research to find real-world governance examples that fill empty spectrum regions. The coverage map drives anchor selection through purposive sampling. Like the library, it is out of the menu until the anchors are re-scored on the seven questions.

A scored three-band record for any figure, institution, or cultural work. It contains scores on each axis for each of the three bands, measured against the anchor library, plus cross-axis findings and band gap analysis. Profiles were scored on the earlier five axes and are out of use until they are re-scored on the seven questions. A profile is the output of Figure Analysis — the figure is measured by the anchors. It may or may not go on to become an anchor itself.

One of three layers in a profile. The universal structure is Individual / Collective / Output, with labels that adapt per entity type. Politicians: Speech / Platform / Governance. Musical groups: Members / Group Positions / Body of Work. Media outlets: Journalists / Editorial Board / Published Coverage. Each band is scored independently. The gap between bands is a finding.

The distance between a figure's scores across the three bands. A narrow gap means consistency. A wide gap reveals the distance between rhetoric and reality. The gap is never averaged away — it IS the finding.

The full evidentiary record for a figure. Contains the raw evidence inventory (organized by source tier, not by which score it supports), rubric-level mappings, adversarial engagement with at least one credible counterargument per score, derived scores with confidence levels, and a full bibliography. The evidence file is evidence-first: all evidence is gathered before any score is assigned.

The evidentiary weight classification for sources used in evidence files. Five tiers: Tier 1 (governance records — highest weight), Tier 2 (primary documents), Tier 3 (peer-reviewed analysis), Tier 4 (quality journalism), Tier 5 (think tank/advocacy — lowest weight, always labeled with known orientation).

A specific reference in the project bibliography with a web link, DOI, or archive location. Points to a source used in an evidence file or in the methodology documentation.

One of 10 types of examination the tool performs: Article, Figure, Anchor, Corpus-Level, Author Trajectory, Outlet Drift, Event Spread, Era-Appropriate Dual Scoring, Score Your Own, and Tilt/Shift. Each type asks a different question and some require accumulated data from other analysis types.

A published finding synthesized from one or more analyses. Written for public audiences. Examples: "The Populism Gap" (synthesized from anchor profiles), "The Content Filter as Overton Finding" (meta-analysis). Timed to events or published on a regular cadence.

A Frequently Waged Accusation entry. A pre-published defense against an anticipated challenge to the methodology or a specific anchor placement. Every entry follows: accusation → wrong question → right question → where it lives in the framework → what the framework actually surfaces. Designed so that every challenge forces engagement with evidence.