Hamtramck Herald
Support the desk

Editorial standards · Methodology

Herald Double-Blind Review Audit™

Independent Multi-Model + Human Editorial Review

The Herald Double-Blind Review Audit is a human-led editorial quality-control method in which multiple independent AI review environments evaluate the same article against a common rubric without receiving the other reviewers' assessments. Their findings are subsequently reconciled and reviewed by a human editor.

This is a newsroom methodology and is not the same as scholarly double-blind peer review. In scholarly publishing, double-blind means authors and reviewers do not know one another's identities. Here it means reviewers do not see one another's assessments.

AI can challenge the journalism; it cannot certify the journalism.

Standing Herald policy. Responsibility stays with the people who publish.

What each reviewer sees

Every reviewer receives

  • the article as drafted
  • the underlying source materials
  • identical evaluation criteria

No reviewer receives

  • another reviewer's score
  • another reviewer's reasoning
  • any desired or target final score

The five stages

  1. 01

    Human assignment

    An editor sets the question, the sourcing standard, and the posture: nonpartisan, cross-partisan, evidence-led. Reporting and drafting are done by people.

  2. 02

    Internal audit, during drafting

    The working draft is scored against the ten-category rubric to find weaknesses. The internal score exists to trigger revision, not to be shown off. Its three largest vulnerabilities are published alongside the article.

  3. 03

    Revision

    The draft is rewritten against those findings — usually for language neutrality, political symmetry, and the separation of fact from interpretation.

  4. 04

    External audit, cold read

    The finished article goes to a fresh review environment that has not seen the internal score, earlier drafts, editorial arguments defending the piece, or any desired outcome. Otherwise the second score is not independent.

  5. 05

    Human reconciliation

    An editor reconciles disagreements, publishes them rather than averaging them away, and makes the publication decision. Where the two totals differ materially, the article explains why.

The nonpartisan article gauge

Ten dimensions, scored 0–10 each, for a total of 100. The Herald publishes all ten category scores rather than a single flashy number — a total without its rubric recreates the opacity the audit exists to solve.

  1. 01

    Factual grounding

    Are material factual claims sourced and verifiable?

  2. 02

    Primary-source reliance

    Are statutes, policies, records, and official documents preferred where available?

  3. 03

    Fact / analysis separation

    Can readers distinguish established facts from the author's interpretation?

  4. 04

    Political symmetry

    Would substantially the same standard apply if the people involved were different?

  5. 05

    Language neutrality

    Does wording avoid loaded praise, insinuation, or campaign rhetoric?

  6. 06

    Counterinterpretation

    Does the article fairly acknowledge reasonable alternative readings?

  7. 07

    Conflict transparency

    Is the author's candidacy and every materially relevant relationship disclosed?

  8. 08

    Legal precision

    Are appointment, election, authority, policy, and statute described accurately?

  9. 09

    Civic educational value

    Does the reader understand the institution better after reading?

  10. 10

    Source traceability

    Can readers readily follow links to the evidence?

Rating bands

  • Exceptional nonpartisan discipline95+
  • Strong90+
  • Publishable with identifiable limitations80+
  • Significant revision recommended70+
  • Do not publish as nonpartisan analysisbelow 70

An article is never manipulated to obtain a preferred score. A score below 70 is not published as nonpartisan analysis.

The AI transparency gauge

A separate 0–100 score measuring how transparently the AI-assisted process is disclosed and documented. It is not a measurement of whether an article is true.

  • Disclosure of AI assistance
  • Identification of human authorship
  • Disclosure of the author's candidacy
  • Source traceability
  • Independent model review
  • Preservation of reviewer disagreement
  • Human final accountability
  • Published correction procedure
  • Absence of fabricated sourcing
  • Reproducibility of the review methodology

The red-team pass

Before publication, the desk argues deliberately against its own article. Any passage that cannot survive these questions is rewritten.

  • Where does this sound like campaign messaging?
  • Where does the author overstate the brilliance of an ordinary policy?
  • Is "peace" being imposed on a routine administrative procedure?
  • Does the article confuse lawful procedure with good procedure?
  • Could a legally valid process still deserve criticism?
  • Does the author's candidacy create an undisclosed strategic benefit?
  • Would the analysis sound different if someone the author opposed had been involved?
  • Are we praising an outcome when the evidence only supports describing a process?

Two final tests close every audit. If every proper noun were removed from the article, would its principles still make sense? And if the people involved were politically or personally opposed to the author, would the same standards apply? If either answer is no, the article is revised again.

Applied continuously to the wire

Every 30 minutes, the wire is swept and each item is scored for nonpartisan character

The Herald's automated sweep pulls regional and cross-border coverage every half hour and runs each headline and summary through an independent model pass against this rubric's nonpartisan tests. The published score describes the item's language and framing only. It is not a verdict on the outlet, and no machine score decides what the Herald publishes.

Read the scored wire

These scores are editorial quality-control indicators created by the Hamtramck Herald. They are not scientific measurements, legal certifications, academic peer review, or guarantees that an article is unbiased. Readers should examine the linked primary sources and reach their own conclusions.

The Hamtramck Herald is part of the Kotcher family of organizations and editorially distinct from it. Kotcher funds the Herald's civic reporting; the Herald endorses no candidate, party, or ballot outcome.