Zacharus Gudmunsen

This thesis aims to establish the possibility of, and a pathway to, artificial moral agents. Artificial moral agents are argued to be of value not just for their practical performance, but because they offer a non-human perspective that can be used to make human theories more objective. The thesis works to a definition of moral agency, arguing that moral agents need to be intentional, morally reasons-responsive, and autonomous, but not necessarily conscious. Then, applying this to artificial agents, it draws on literature from moral epistemology and responsibility to argue that artificial agents normally fail to meet these criteria because they are not simultaneously morally reasons-responsive and autonomous. Following this, it argues that the most promising means of developing artificial moral agents is for artificial agents to evolve into moral agents. Even if not moral agents precisely, the evolutionary development suggested seems likely to produce autonomous artificial agents that respond to some moral reasons, which would still offer the desired non-human perspective.

Raul Hakli & Pekka Mäkelä (2016, 2019) make a popular assumption in machine ethics explicit by arguing that artificial agents cannot be responsible because they are designed. Designed agents, they think, are analogous to manipulated humans and therefore not meaningfully in control of their actions. Contrary to this, I argue that under all mainstream theories of responsibility, designed agents can be responsible. To do so, I identify the closest parallel discussion in the literature on responsibility and free will, which concerns ‘design cases’. Design cases are theoretical examples of agents that appear to lack responsibility because they were designed, philosophers use these cases to explore the relationship between design and responsibility. This paper presents several replies to design cases from the responsibility literature and uses those replies to situate the corresponding positions on the design and responsibility of artificial agents in machine ethics. I argue that each reply can support the design of responsible agents. However, each reply also entails different levels of severity in the constraints for the design of responsible agents. I offer a brief discussion of the nature of those constraints, highlighting the challenges respective to each reply. I conclude that designing responsible agents is possible, with the caveat that the difficulty of doing so will vary according to one’s favoured reply to design cases.

Humans are responsible moral agents in part because they can competently respond to moral reasons. Several philosophers have argued that artificial agents cannot do this and therefore cannot be responsible moral agents. I present a counterexample to these arguments: the ‘Moral Decision Machine’. I argue that the ‘Moral Decision Machine’ responds to moral reasons just as competently as humans do. However, I suggest that, while a hopeful development, this does not warrant strong optimism about ‘artificial moral agency’. The ‘Moral Decision Machine’ (and similar agents) can only respond to moral reasons by deferring to others, and there are good reasons to think this is incompatible with responsible moral agency. While the challenge to artificial moral agency based on moral reasons-responsiveness can be satisfactorily addressed; the challenge based on moral deference remains an open question. The right way to understand the challenge, I argue, is as a route to the claim that artificial agents are unlikely to be responsible moral agents because they cannot be authentic.