Work

Structured Disagreement as an Architecture for Multi-Model AI

Year: 2026

Type: article

Abstract: This article presents xCheckAi, a structured, multi-model comparative framework designed to resist oracle-like use of large language models in consequential decision contexts. Rather than producing a single synthesised answer or enforcing consensus, the framework exposes agreement, disagreement, uncertainty, and refusal across independently queried models, preserving provenance and auditability at each stage. The system is intentionally non-prescriptive and does not determine truth or recommend actions; instead, it functions as epistemic infrastructure aimed at preserving human responsibility and preventing the quiet transfer of judgement to AI systems.

Source: Zenodo (CERN European Organization for Nuclear Research)

DOI: 10.5281/zenodo.18588194

Author: Harrison, Kyle Antony

Language: English

Topic: Ethics and Social Impacts of AI

Subfield: Safety Research

Field: Social Sciences

Domain: Social Sciences

SDG: Peace, justice, and strong institutions

Open Access status: green

OpenAlex ID: W7128484452

FWCI:

Cites:

Cited by:

Related to: