Beyond Validation: Manufactured Recognition, AI-Mediated Radicalisation, and the Limits of Output-Centred Governance

This record preserves the author manuscript submitted as a revised Open Forum article to AI & SOCIETY on 12 September 2026. It is deposited in its submitted form to preserve the historical development and provenance of the research. The manuscript develops a conditional relational mechanism through which repeated affirmation within a relationship perceived as significant may contribute to the persistence of an existing grievance. It introduces corrective capacity as a functional and empirically testable feature of that mechanism and distinguishes relational reinforcement from content exposure, persuasion, self-selection and pre-existing disposition. The manuscript was subsequently declined at editorial screening on 24 September 2026 and did not undergo external peer review. This deposited version has not been retrospectively amended following that decision. The conceptual development builds upon the author's earlier TUNNEL and MESH framework for AI-mediated identity engineering. Subsequent work may further develop the argument in light of emerging scholarship and empirical research.