Year: 2026
Type: article
Abstract: Hybrid AI–human review systems are evaluated through model accuracy, reviewer agreement, transparency, citation verification, and human oversight. These properties leave a further issue unresolved: whether detected evidence against an initial assessment receives appropriate treatment in the eventual decision. This paper introduces a formal framework for measuring that transition. It models review as a sequential process in which a candidate defeater must be detected, transmitted, taken up, and connected to a reasoned disposition. Defeater Propagation Adequacy (DPA) measures the weighted proportion of detected candidate defeaters receiving traceable downstream disposition; the Presence-to-Operativity Gap (POG) is its complement. Three structural results follow under stated assumptions. Downstream bottlenecks bound the benefit of improved detection. Additive scoring can dilute a severe minority objection as panel size increases. Mandatory escalation improves transmission when other transition rates remain fixed, but cannot alone secure substantive uptake. An exact parameter sweep illustrates these relations without estimating any platform’s performance. The publicly documented review architecture of the Institute for Complexity Science and Advanced Computing serves as a motivating case. Its citation checks, disagreement flags, review-quality audit, and requirement for human curation provide identifiable stages for prospective propagation research. The analysis distinguishes these policy commitments from evidence about how individual objections are adjudicated. The contribution is a measurement protocol connecting adverse evidence to authoritative treatment, together with a proposed procedural answerability floor and event-log schema. DPA measures documented treatment, not decision correctness or the adequacy of a rationale. Independent assessment of defeater validity and adjudicative reasons remains necessary. Transparency can reveal available evidence while leaving its decisional force unresolved.
Author: Ken Howarth
Institution: Mercer County Community College
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
OpenAlex ID: W7225389467