From Inference to Institutional Judgment: Governing AI-Mediated Representations About People

Abstract

Institutions increasingly act through profiles, classifications, summaries, predictions, and generated accounts about people. Existing privacy and data-governance theories explain many problems concerning information flows, prediction, group effects, and epistemic injustice. A remaining problem is how a representation becomes an operative premise of institutional judgment and how its effects can be repaired after it has spread. This article develops a theoretical pathway from information access and representation through inference, persistence, institutional availability, material reliance, consequence, challenge, and downstream repair. A continuing hypothetical housing-profile case translates the pathway without replacing its full provenance and relational structure. The framework distinguishes access from power, inference from knowledge, accuracy from justification, availability from reliance, and source correction from downstream remedy. Relational epistemic power is used only where an actor or institutional arrangement can shape representations in ways that affect another party’s treatment or ability to contest. Two figures trace the representation to-consequence and repair pathways. The article is theoretical and reports no original dataset, product audit, participant study, systematic review, or empirical validation.