Purpose

This statement defines how Undeterred Press verifies that its editorial review process functions as described publicly: AI-assisted triage followed by a genuine, independent human review, with the human reviewer holding sole authority over the accept, revise, or reject decision. It exists so that "editor-reviewed" remains a claim the journal can substantiate, not just assert — to itself as much as to readers.

The terminology used throughout this statement — "editor-reviewed," the roles and stages described below — is informed by ANSI/NISO Z39.106-2023, a terminology standard published by NISO in 2023 to give publishers a shared, comparable vocabulary for describing review models. We're a small outlet without the resources to seek formal accreditation against it, but using its terminology deliberately, rather than inventing our own in a vacuum, is a small way of describing our process in language the field already recognizes.

What Is Being Assured

Not manuscript quality. Manuscript quality is the reviewer's judgment call, and judgment calls vary — that's expected and healthy. What this statement assures is process integrity: that a real, independent human read of the full submission occurred before every decision, and that the AI triage brief functioned as an aid to that read rather than a substitute for it.

The AI Triage Boundary

AI-assisted triage may:

AI-assisted triage may not, under any circumstance:

Reviewer Requirements

Every review decision, regardless of reviewer, is accompanied by a completed decision form requiring:

  1. Two specific citations — one passage representing the argument's strongest point, one representing its weakest or most vulnerable point, quoted or closely paraphrased from the submission itself.
  2. An independent restatement of the core claim and methodology, in the reviewer's own words, not lifted from the triage brief's framing.
  3. A decision rationale referencing something specific to the piece — a figure, a finding, a phrase — rather than language generic enough to apply to any submission.
  4. Verification of the manuscript's most load-bearing citations — do the sources actually say what the manuscript claims they say. Where a claim shows meaningful variance across the author's or a cited source's own prior framings of the same underlying evidence (a blog post stating something more strongly than the peer-reviewed version of the same finding, for instance), that variance is a trigger to check the underlying primary evidence directly — not to default to whichever version reads as more cautious. The more hedged framing isn't automatically the correct one; it's just the other framing.

A review missing any of these elements is incomplete and the decision does not stand until it's provided.

Self-Review Discipline

For categories where the Founding Editor is also the reviewer, an added safeguard applies to keep judgment independent of the AI-generated brief rather than anchored to its framing. The specifics of that safeguard are an internal editorial practice; the commitment that matters publicly is that the reviewer's own read of the full submission, not the brief, is what the decision rests on.

Spot Checks

On an ongoing basis, the editor independently re-reads a sample of past decisions, blind to the original reviewer's conclusion, and compares. Disagreement is expected and useful — it calibrates reviewer standards over time. A pattern of disagreement, or a re-read that contradicts a decision's stated rationale in a way the citations don't support, is the signal worth acting on.

If a Lapse Is Found

If a spot audit or timing flag reveals a decision wasn't grounded in a genuine independent read:

Review of This Statement

This document is revisited whenever reviewer roster size changes, submission volume changes materially, or a lapse under the section above reveals a gap the current process doesn't cover.

Companion document: see the Editorial Triage & Review Workflow for the step-by-step path a submission actually takes.