Unadjudicable Error: Infodebt, Technical Adjudication, and the Hollowing of Assurance

Machine-generated outputs increasingly function as reasons for belief, decision, and action. A classifier denies credit, a logger record releases a temperature-sensitive consignment, a certificate licenses a cladding system, an amended type certificate carries a derivative aircraft into service. In each case an output acquires consequential force. This book argues that consequential force is not adjudicative settlement, and that much of what contemporary governance treats as assurance is evidence of the first offered as though it established the second. The argument proceeds at three levels. At the level of the claim, it introduces Infodebt: the conditional adjudicative liability opened when a socio-technical practice authorizes downstream dependence on an output before any judgment of that dependence has been performed. Claim-standing is used in a role-functional sense that presupposes no machine speaker, believer, or bearer of intentional states. A constitutive argument then identifies two minimal requirements of technical adjudication: an observational background that makes the claim's grounds and applicability answerable, and a judgment-outcome space in which assessment can make an authorized difference to the dependency at issue. A realization condition requires that the disposition attributed to an assigned adjudicative position actually be formed through assessment of the relevant grounds. From it follows a non-proxy principle: evidence, explanations, certificates, and prior judgments may be inherited, but propagation cannot substitute for performance of a newly required adjudicative operation. At the level of governance practice, a deletion test applied to the oversight provisions of the EU Artificial Intelligence Act and the General Data Protection Regulation shows that the two requirements are non-substitutable. Without answerability the residual relation is unanswerable discretion; without consequential response it is advisory review. A third, harder failure—joint-exercise failure—arises when both are formally present but the attributed assessment does not participate in forming the disposition. Because records establish attribution rather than participation, evidence of safeguard presence is non-additive: it cannot by itself establish that oversight occurred. At the level of institutional dynamics, structural hollowing explains how a qualification can persist while the realization it stands for declines, without deception or rule-breaking. Adjudicability is treated as a maintained institutional capacity rather than a fixed monitoring parameter. Acceptance sets can be thick because task-relevant differences are never adjudicated or because adjudicated differences are normatively ignored; where qualification has independent value, repeated selection and hysteresis turn local economization into a ratchet. The assessment of external wall fire performance in England before 2018 and the certification of the Boeing 737 MAX illustrate the mechanism. The framework is deliberately typed. It establishes necessary conditions of performed technical adjudication, not sufficient conditions for truth, reliability, fairness, or legitimacy, and it declares the points—trigger thresholds, individuation, the authority of standards—at which it must return ‘cannot determine’. Its unifying thesis, stated for the structural dimension of trustworthiness the book examines, is that the opposite of trustworthiness is not error but unadjudicable error: error that, under a given institution, standard, and time, cannot become an object of evidence, denial, or revision.