When Does Automated Review Constitute Genuine Reconsideration?

An institution can automate both an initial decision and its review without making the original determination genuinely vulnerable to challenge. This paper examines what distinguishes genuine institutional reconsideration from a second computational procedure that merely reproduces or rationalizes an existing decision. It develops the Criterion of Non-Redundant Defeatability, arguing that, within a constituted institutional relation of commitment and challenge, an automated reviewing architecture possesses the capacity for genuine reconsideration if and only if it satisfies three conditions: Evidential Independence (E): The capacity to examine independently assessable grounds rather than merely inherit the original decision’s justification. Adversarial Defeatability (D): Both structural permission and effective competence to recognize, assess, and pursue material reasons capable of defeating the institution’s maintained position. Normative Effectiveness (N): An authorized route through which the resulting assessment can govern the institution’s operative position. Drawing on defeasible reasoning, discursive commitment, epistemic reliance, and scholarship on responsibility gaps, the paper distinguishes the institution as the bearer of commitments from the automated reviewer as an epistemic delegate. It uses contrast cases and a credit-review example to examine the necessity and sufficiency of the proposed conditions, addresses the problem of evaluating defeater competence without circularity, and identifies evidential requirements for auditing reviewing architectures. A central contribution is the distinction between possessing the capacity for reconsideration, adequately exercising that capacity, and discharging the resulting institutional obligations. The account explains why technical independence, explanatory fluency, human approval, procedural completion, and even correct outcomes are not, individually, sufficient to establish genuine reconsideration. The resulting framework provides a conceptual basis for evaluating automated administrative review, algorithmic contestability, and institutional answerability in legal and public-sector decision-making.