An Analysis of the Possibility of AI Moral Agency Based on the Origins of Voluntary Action in Transcendent Philosophy

Recent advances in artificial intelligence (AI) have intensified debate over whether systems capable of learning, information processing, output selection, operational goal pursuit, and linguistic interaction can qualify as moral agents. This theoretical study employs philosophical analysis to reconstruct the origins of voluntary action in Transcendent Philosophy and assess whether contemporary AI satisfies the conditions required for moral agency in the sense relevant to moral responsibility. These origins form an ordered structure rather than a set of independent capacities: intention presupposes knowledge of the end; will presupposes the end’s desirability and an internal motive; and free choice concerns the conscious issuance of action from the agent through this structure. The analysis then examines the affective and normative capacities required for moral answerability and situates AI within the Sadrian modes of agency. It argues that information processing does not amount to knowledge merely by becoming available for computation; objective functions and reward signals do not show that the system knows and desires an end; and algorithmic selection does not amount to voluntary choice. Nor do emotionally expressive language and norm-sensitive responses suffice to demonstrate affective experience or normative understanding. Yet denying moral agency need not reduce AI to a wholly passive instrument. By analytically extending the concept of the agent by subjection (al-fāʿil bi-l-taskhīr), the paper argues that contemporary AI may exercise genuine causal and operational agency while lacking independent moral agency. This conclusion concerns contemporary material-computational systems and does not establish the impossibility of artificial moral agency as such.