Should we speak of machine agency? A case against conceptual extension

Abstract An ever-increasing number of digital technologies are attributed capabilities of the kind that were so far typically considered unique to humans. This raises many widely discussed ethical issues, but more intimately, this development questions the way we should relate to these technologies, and in which terms we should talk about them. This article introduces the notion of conceptual extension to describe the use of a concept that was rather reserved for humans, to machines. It builds on recent work on pragmatic conceptual ethics to argue that the acceptability of conceptual extension is primarily a normative question: instead of examining whether a machine fulfills the requirements of a given definition, we should ask what function the application of a given concept to machines would fulfill. Exemplifying this approach on the concept of agency, the article reviews different positions on machine agency and discusses the normative commitments to serve different functions that underlie positions for or against the extension of agency to machines.