Moral Responsibility of Robots and Hybrid Agents

The Monist, 2019, 102, 259–275 doi: 10.1093/monist/onz009 Article Moral Responsibility of Robots and Hybrid Agents ABSTRACT We study whether robots can satisfy the conditions of an agent fit to be held morally responsible, with a focus on autonomy and self-control. An analogy between robots and human groups enables us to modify arguments concerning collective responsibility for studying questions of robot responsibility. We employ Mele’s history-sensitive account of autonomy and responsibility to argue that even if robots were to have all the capacities required of moral agency, their history would deprive them from autonomy in a responsibility-undermining way. We will also study whether humans and technological artifacts like robots can form hybrid collective agents that could be morally responsible for their actions and give an argument against such a possibility. 1. INTRODUCTION As a consequence of increasing automation in our societies, algorithms and robots are operating and making decisions in areas that used to be controlled by humans alone, for instance, in stock trading, medical diagnosing, and car driving. The question has been raised what happens to responsibility when human beings give away control (see, e.g., Gunkel [2017] for recent discussion). Traditionally, at least since Aristotle (1985 III.1–5), responsibility has been closely related to both control and knowledge: in order for an agent to be responsible for their actions, they must be in control over those actions and aware of what they are doing. When humans create and employ robots and algorithms that replace other human beings in decision making, they lose control and also give up knowledge of what kinds of decisions are made, because the decisions are based on artificial intelligence (AI) and machine learning techniques, whose behavior can sometimes be unpredictable. Even the programmers themselves are often surprised how their programs behave after the learning phase. Often, humans do not even know what kinds of actions will be performed and what their consequences might be. It thus seems that both traditional conditions of responsibility—control and knowledge—are out of reach for human beings, including the designers, builders, controllers, owners, and users of those machines. *RH: University of Helsinki, PM: University of Helsinki C The Author(s), 2019. Published by Oxford University Press. V All rights reserved. For permissions, please email:  259 Raul Hakli and Pekka M€akel€a* 260  Moral Responsibility of Robots and Hybrid Agents What makes things worse is that even thinking in terms of collective or shared responsibility does not seem to help much: Just as there is no single individual who possesses the necessary knowledge and control of the algorithmic decisions, there is no collection of human beings that would possess the necessary knowledge and control either. Hence, it seems, no one can be held morally responsible for the resulting outcomes. There is a growing body of literature discussing this issue as the problem of responsibility gaps (see Matthias 2004; Gunkel 2012; Gunkel 2017). Traditionally responsibility gaps have been discussed in the context of collective action: When an untoward event happens as a consequence of the actions of multiple individuals, none of whom was in a position to foresee the untoward consequences, it seems that there is no one who can be held responsible. Hence, it has been suggested that perhaps the collective consisting of these individuals could be held responsible. When the use of technology is added to the picture, the situation becomes even more complex. The development of technology already involves many hands, its regulation and use involves further agents, and the technology itself can be so complex that its behavior in new circumstances cannot be reliably predicted. Hence the question arises, who can be held responsible for untoward effects. Some people have argued that perhaps in the case of autonomously operating intelligent systems, it would be possible to attribute the responsibility to these systems themselves (see, e.g., Sullins [2006]). However, it is not easy to see how that could be done. It is not even easy to understand what it could mean in practice to hold a robot responsible. The problem is that even if we were to admit that machines are agents and capable of possessing responsibility-relevant control and knowledge of their actions, they do not seem to qualify as moral agents and thus they do not seem to satisfy the general conditions of agents fit to be held responsible for their actions. Often mentioned candidates for necessary conditions of moral agency are personhood, autonomy, self-awareness or consciousness, sensitivity to moral reasons, capability to experience emotions, etc. There is no consensus on what the exact conditions of moral agency are, and there is an ongoing discussion of how high the bar is set by the notion of moral agency presupposed by our moral practices (see, e.g., McKenna [2006]). Despite the lack of agreement concerning what the exact conditions of moral agency are, the majority view is that present-day robots and AI agents do not satisfy them. An agent must satisfy the conditions of moral agency in order to qualify as a proper bearer of moral responsibility. Hence, according to this line of thinking, current robots and AI agents are not fit to be held morally responsible. In this paper we are not only interested in present-day robots but on the conceptual possibility of attributing moral responsibility to robots. We will argue that robots are not the types of agents that could appropriately be held morally responsible because as products of engineering they cannot be autonomous in the sense of autonomy relevant for moral responsibility. However, some philosophers have argued that although artificial agents like robots were not full moral agents and hence not capable of bearing full responsibility, they still modulate people’s behavior in such a way that they could be held partially responsible for the outcomes of those actions (see, e.g., Hanson [2009]). Human Moral Responsibility of Robots and Hybrid Agents  261 beings and technological artifacts acting in interaction can form so-called hybrid agents, which could be seen as collective agents satisfying the conditions of moral agency. As relatively autonomous parts of these hybrid agents, the artificial agents in them would then carry some responsibility for their part of the actions in the actions of the collective agent (Neuh€auser 2015). In contrast to these views, we will defend the conditional claim that if robots cannot be moral agents, as argued here, then they cannot be partially responsible either. One reason for being concerned with the tendency to extend usage of normative notions like autonomy, moral agency, and moral responsibility to artificial agents like robots is that such notions play a constitutive role in imp (...truncated)