Corrective Ethics Under Constraint: Mediated Authority, Answerability, Burden, and Correction under Finite Action

This paper argues that ethical evaluation of institutions and systems must include not only outcomes, procedures, and formal rules, but also the mediated authority conditions through which cases become governable under finite constraint. Institutions do not act on cases in exhaustive presence, but through selective operative representations—files, charts, scores, categories, thresh- olds, and model outputs—that preserve some structure strongly enough for action while leaving other structure weakly carried, backgrounded, merged, or omitted. Its central claim is that once consequence-bearing authority governs through such renderings, the mediated pathway through which action becomes possible is no longer merely technical background. Because authority reaches the case only through the rendering, defects in that rendering become defects in the act’s mode of address to its object, and are therefore ethically assessable. This matters especially where the governed case is a humanly lived case, or subject-process, whose lived significance is adequacy-relevant. The paper argues that operative representations are interface-shaped, purpose-bound, level-bound, and scope-limited, and that they remain structurally non-identical to the cases they govern. Once authority acts through them, omissions, burden displacements, scope overreach, experiential distortion, and blocked correction partly constitute the mediated act itself. The paper therefore identifies mediated authority as a distinct middle-layer moral object, uses answerability as the bridge from architecture to ethics, and derives lower-bound constraints of non-totalization, scope discipline, traceability, burden sensitivity, experiential responsiveness, and correction responsibility. It is not a complete morality, legitimacy theory, or politics of institutions, but a structural ethical account of how finite systems govern through selective renderings of cases.