Agency Attributions and AI Agency

The contemporary debate surrounding AI agency is frequently stalled by an essentialist preoccupation: attempting to discern whether an artificial system possesses genuine autonomy or traditional belief-desire agency. This paper argues that such an approach is a category error that obscures the practical realities of AI deployment. By shifting the discourse from an ontological investigation of machine agency to an “attributionist” framework, we can better understand how agency is constructed through socio-technical relationships rather than residing solely in algorithmic code. Drawing upon the theological concept of fā‘il bi-l-tajallī – agency expressed through mediated manifestation – this paper posits that AI systems act as conduits for human intent and institutional structure. This perspective provides a necessary alternative to current technical debates. Furthermore, by engaging with recent scholarship on responsibility gaps – specifically by synthesizing Sven Nyholm’s seminal identification of the moral vacuum in autonomous systems with Herman Veluwenkamp’s granular analytic distinctions – we move toward a more precise taxonomy of accountability. Veluwenkamp shows that the so-called “responsibility gap” must be disaggregated into Epistemic Responsibility Gaps (ERG), relating to system complexity and explainability, and Control Misalignments (CM), which stem from institutional delegation failures. Ultimately, this paper reframes the “Alignment Problem.” It is not a purely technical challenge of matching an objective function to a static goal, but a socio-technical endeavor of normative integration. An aligned AI is not one that achieves perfect independence, but one whose operations remain within the legible, revisable, and accountable bounds of human institutional intent. Through this attributionist lens, we demonstrate that resolving alignment concerns requires less focus on the internal “will” of the machine and greater focus on the distributed responsibilities that arise with the instantiation of these systems.

کلیدواژه ها:

agency ، AI agent ، reflex ، action ، Autonomy ، Responsibility Gap ، agent by self-disclosure