Do Artificial Agents Have Free Will?

With the new developments in artificial intelligence research, artificial agents (AAs) become more and more autonomous, thus it is natural to suspect that they might, to certain extent, have free will. However, according to the mainstream view in traditional philosophy as well as to most people’s intuition, it is impossible to attribute free will to any machine. This paper engages in this debate by developing a concept of “practical freedom” on the basis of the classic “reactive attitudes” theory. It argues that, while AAs may satisfy two important conditions required by the concept of practical freedom, i.e., the unpredictability of behavior and the potentially conflicting goals, they can hardly satisfy the key condition of mutual recognition of expectations and requirements, thus are unlikely to be regarded as having freedom of will in the near future. Through the analysis of the possible conditions of the freedom of will of AAs, this paper also aims to deepen our understanding of the nature and foundation of the freedom of will (in humans).