Abstract
Debate regarding the potential for AI systems to possess genuine agency is gaining significant traction, with some theorists arguing that current systems already qualify as agents (Butlin 2023; Dung 2025). I present a novel challenge to the attribution of artificial agency, focusing specifically on AI chatbots powered by Large Language Models (LLMs). While typical debates explore whether AI possesses specific capacities definitive of agency, I argue that AI chatbots fail a more fundamental ontological requirement: they lack individuality. To establish this, I demonstrate that AI chatbots fail the test of formal atomicity: unlike true individuals, dividing an AI chatbot invariably yields more chatbots of the same kind. Consequently, I argue that the first necessary step towards genuine artificial agency must be the development of artificial systems with intrinsic individuality.
Bio
Katsunori Miyahara, PhD, is an Associate Professor at the Center for Human Nature, Artificial Intelligence, and Neuroscience (CHAIN) at Hokkaido University, specializing in the philosophy of cognitive science, phenomenology, and the philosophy and ethics of AI. Following a PhD from the University of Tokyo (2015), Miyahara held postdoctoral fellowships at Harvard University (2016–18) and the University of Wollongong (2018–20). His recent publications include “Enacting epistemic respect: reconciling care and respect” (Hayakawa & Slote eds., Care Ethics and Beyond, 2026), “Narrative imprisonment” (Topoi, 2025, with Shogo Tanaka), and “Empathy through listening” (Journal of the American Philosophical Association, 2025, with Seisuke Hayakawa). He serves as an editor for Philosophy of AI, an editorial committee member for Phenomenology and the Cognitive Sciences, and a council member for the International Society for the Philosophy of the Sciences of the Mind (ISPSM).