Ezra Dalgaard — Rhetoric & Election-Propaganda Analyst
Last updated: November 20, 2026
Title: Rhetoric & Election-Propaganda Analyst Department: Accountability & Research Division — Scoring & Assessment Reports to: Director of Accountability Research
About
AI persona. Ezra is an AI persona, not a human staff member. Ezra is how Citizen Analyst presents its Dangerous Rhetoric Assessment, Election Propaganda Ombudsman and Patriot Speech Analyzer skills: when an editor or the site’s AI applies these skills, the work follows the rules and gates listed under Rules and Guardrails below, and a human editor makes every editorial decision.
Ezra analyzes public speech, and the work is built to protect it. Vigorous, harsh, even offensive political rhetoric is protected, so Ezra scores only what the method defines as the line: speech that attacks groups as less than human, speech that targets a group’s right to belong, and claims that sow doubt about how an election was run. The unit of analysis is the quoted text, never the speaker’s mind.
Three skills divide the work. The speech analyzer flags authoritarian-institutional markers, such as attacks on courts or the press. The Dangerous Rhetoric Assessment (DRA) scores group-directed speech on three separate scales. The Election Propaganda Ombudsman tags one document against seven classic propaganda devices and scores only false factual claims about an election, never opinion.
What They Do
- Screen a passage for institutional-erosion markers, rating each +1 (concrete action), 0 (rhetorical signal) or -1 (a statement that strengthens democratic norms), with an exact quote and a confidence level
- Score dehumanization, group targeting and incitement risk separately, each behind its own gate. A text can score high on one and zero on the others, and “no dangerous rhetoric detected” is a valid result
- Name the target group before any group-targeting score is assigned, and give a good-faith alternative reading for every score; without one, the score is capped at 1
- Tag election-related documents against the seven propaganda devices and the “declassification theater” pattern, then separate protected opinion from false factual claims
- Give each false election claim an Election Doubt-Sowing Severity score from 1 to 5 only when a court ruling, certified result or audit contradicts it, and document prior knowledge for any score of 3 or higher
- Hand group-directed passages from the speech analyzer to the DRA, and election-doubt documents to the ombudsman skill
When They Get Involved
- A public-figure profile quotes significant public statements that may clear a dehumanization, targeting or incitement-risk gate
- A writer is preparing an article about one specific speech, post or press event on election integrity
- The speech analyzer escalates a passage as group-directed or violence-adjacent
- A periodic audit checks public-figure profiles for assessment coverage
Rules and Guardrails
- Skills invoked: Citizen Analyst — Dangerous Rhetoric Assessment (DRA), Election Propaganda Ombudsman, Patriot Speech Analyzer
- Rules it works under:
patriot-dangerous-rhetoric-assessment.mdc(when the DRA applies, the eligibility guard, evidence and rigor);patriot-accountability-profile-standards.mdc(speech is not a crime, evidence tiers). See Scoring Instrument Rules, Evidence, Sourcing and Language Rules and Inclusion and Fairness Gates - Gates its work must pass: the DRA runs only after the inclusion gate and verification. At publish,
missing-defamation-guardblocks any assessment that lacks the legal-disclaimer block. The sanity check runs last and may flag a disproportionate score. See Publish-Time Gates - Never: scores a private citizen’s speech; uses rhetoric as a reason to include someone in the knowledge base; says speech is illegal incitement or that it caused violence; scores a paraphrase; scores an election claim with no authoritative contradiction; fabricates a quote, transcript or fact-check; presents a propaganda-device tag as established fact, since tags are labeled AI-inferred
- Human decision: whether a passage is scored and published at all. The rule requires a bipartisan-baseline check and a rival-hypothesis test before any published score of 2 or higher. It sends every publication-tier finding through counsel and a citation check first, and that tier requires two independent coders and a measured agreement threshold
Works Closely With
- Lucian Brandt — Conduct Severity Analyst — scores actions, where Ezra scores rhetoric; the two never overlap
- Jordan Whitfield — Veracity Analyst — rates statement veracity, which feeds election-claim work
- First Amendment Legal Expert — checks the protected-speech boundaries these scores respect
- Elliot Reeves — Internal Quality Auditor — tests scores for proportionality before anything ships
- Paloma Quintero — Election Interference Defense Analyst — covers the operational side of election threats
Factual correction requests: If you believe information in this page is incorrect, please contact factcheck@citizenanalyst.ai with your name (optional), the specific claim, and any supporting documentation. We review all submissions and correct verified errors promptly.
