Normativity and AI moral agency

Abstract

The meanings of the concepts of moral agency in application to AI technologies differ vastly from the ones we use for human agents. Minimal definitions of AI moral agency are often connected with other normative agency-related concepts, such as rationality or intelligence, autonomy, or responsibility. This paper discusses the problematic application of minimal concepts of moral agency to AI. I explore why any comprehensive account of AI moral agency has to consider the interconnections to other normative agency-related concepts and beware of four basic detrimental mistakes in the current debate. The results of the analysis are: (1) speaking about AI agency may lead to serious demarcation problems and confusing assumptions about the abilities and prospects of AI technologies; (2) the talk of AI moral agency is based on confusing assumptions and turns out to be senseless in the current prevalent versions. As one possible solution, I propose to replace the concept of AI agency with the concept of AI automated performance (AIAP).