Resolutionist AI — Agency and Responsibility in Artificial Systems

Contemporary AI ethics is dominated by the concept of alignment: the project of ensuring that artificial systems reliably pursue human-specified goals. This paper argues that alignment is conceptually inadequate for grounding agency or responsibility. Treating aligned behaviour as a basis for responsibility commits a category error by conflating optimisation with authorship. The paper develops Resolutionism, an account of agency that locates authorship in the act of resolving deliberation into owned commitment rather than in behavioural sophistication, motivational strength, or causal indeterminacy. On this view, artificial systems may evaluate, optimise, and revise goals, but they do not resolve deliberation in the sense required for moral agency. The paper further introduces the Functional Responsibility Threshold, distinguishing moral agency from regulatory responsibility. While advanced systems may warrant functional responsibility for governance purposes, moral responsibility remains anchored in the human decisions that design, authorise, and deploy them. Resolutionism thus provides a structural foundation for AI ethics that preserves accountability without anthropomorphism.