When responsibility cannot arise: algorithmic mediation and the preconditions of moral imputability
This paper argues that moral responsibility can fail under technologically mediated conditions in two structurally con-nected ways. First, agents may act from motivational states significantly shaped by external formative processes whoseinfluence remains partially opaque to reflective awareness. Second, institutions may assign formal responsibility for deci-sions whose operative reasons agents cannot meaningfully access, evaluate, or contest within ordinary practices of delib-eration. The first failure condition is developed as the Conditional Reflexive Requirement (CRR), according to whichblameworthiness depends in part on an agent’s reasonable capacity to identify and critically reassess formative influencesshaping action-guiding motivations. The second is described as misfired imputation, referring to cases in which institu-tions assign responsibility under the assumption that the epistemic and deliberative conditions of genuine agency remainintact when they are structurally absent. The paper argues that these are not two adjacent problems but two moments ofone transformation, and defends that claim through a worked case rather than by analogy: in algorithmically mediateddecision environments a single architectural property, the placement of the operative determinants of conduct outside thepractice of giving and asking for reasons, produces both failures, and the individual-level failure is itself a cause of theinstitutional one. Three theses are defended: common source, causal coupling, and non-separability of remedy. The lastis what makes the two conditions one phenomenon, and it is falsifiable. The argument is developed through a TrackingPrinciple governing responsibility and motivational ownership, a three-part taxonomy of motivational opacity, a thresholdaccount distinguishing the scalar property of contestability from the threshold predicate of misfired imputation, and aconceptual distinction between misfired imputation and adjacent failure modes including the responsibility gap, distrib-uted responsibility, and the accountability deficit. The paper closes by bringing the result to bear on the human-oversightrequirement of Article 14 of the EU AI Act.
Published in AI and Ethics, 2026
Read the article here