When Does Automated Review Constitute Genuine Reconsideration?

An institution can automate both a decision and its appeal while insulating the initial determination from challenge. What distinguishes genuine reconsideration from a second computational performance? This article develops the Criterion of Non-Redundant Defeatability: within an independently identified institutional relation of commitment and challenge, a reviewing architecture has the capacity for genuine reconsideration if and only if it possesses Evidential Independence, Adversarial Defeatability, and Normative Effectiveness. It must examine independently assessable grounds, permit and competently pursue material defeaters, and connect assessment to an authorized route capable of governing the maintained position. Drawing on defeasible reasoning, discursive commitment, epistemic reliance, and responsibility-gap scholarship, the argument distinguishes the institution as commitment bearer from the automated reviewer as epistemic delegate. Contrast cases establish individual necessity; a credit-review example tests whether full computational transparency is a further constituent. The article specifies what would refute sufficiency, defends the relational scope conditions, and addresses the difficulty of assessing competence without presupposing the reviewer’s verdicts. It also distinguishes architectural capacity from actual reconsideration and discharge of an obligation. The resulting account explains how fallible, finite review can reopen institutional warrant without making technical diversity, fluent explanations, human sign-off, or administrative closure sufficient for answerability.