As AI Systems become increasingly autonomous, they are expected to engage in decision-making processes that have moral implications. In this research we integrate theoretical and empirical lines of thought to address the matters of moral reasoning and moral uncertainty in AI Systems. We reconceptualize the metanormative framework for decision-making under moral uncertainty and we operationalize it through a latent class choice model. The core idea being that moral heterogeneity in society can be codified in terms of a small number of (...) classes with distinct moral preferences and that this codification can be used to express moral uncertainty of an AI. Choice analysis allows for the identification of classes and their moral preferences based on observed choice data. Our reformulation of the metanormative framework is theory-rooted and practical in the sense that it avoids runtime issues in real time applications. To illustrate our approach we conceptualize a society in which AI Systems are in charge of making policy choices. While one of the systems uses a baseline morally certain model, the other uses a morally uncertain model. We highlight cases in which the AI Systems disagree about the policy to be chosen, thus illustrating the need to capture moral uncertainty in AI systems. (shrink)

In this paper, I will argue that automated vehicles should not swerve to avoid a person or vehicle in its path, unless they can do so without imposing risks onto others. I will argue that this is the conclusion that we should reach even if we start by assuming that we should divert the trolley in the standard trolley case. In defence of this claim, I appeal to the distribution of moral and legal responsibilities, highlighting the importance of safe spaces, (...) and arguing in favour of constraints on what can be done to minimise casualties. My arguments draw on the methodology associated with the trolley problem. As such, this paper also defends this methodology, highlighting a number of ways in which authors misunderstand and misrepresent the trolley problem. For example, the ‘trolley problem’ is not the ‘name given by philosophers to classic examples of unavoidable crash scenarios, historically involving runaway trolleys’, as Millar suggests, and trolley cases should not be compared with ‘model building in the sciences’, as Gogoll and Müller suggest. Trolley cases have more in common with lab experiments than model building, and the problem referred to in the trolley problem is not the problem of deciding what to do in any one case. Rather, it refers to the problem of explaining what appear to be conflicting intuitions when we consider two cases together. The problem, for example, could be: how do we justify the claim that automated vehicles should not swerve even if we accept the claim that we should divert the trolley in an apparently similar trolley case? (shrink)

Crashes involving self-driving cars at least superficially resemble trolley dilemmas. This article discusses what lessons machine ethicists working on the ethics of self-driving cars can learn from trolleyology. The article proceeds by providing an account of the trolley problem as a paradox and by distinguishing two types of solutions to the trolley problem. According to an optimistic solution, our case intuitions about trolley dilemmas are responding to morally relevant differences. The pessimistic solution denies that this is the case. An optimistic (...) solution would yield first-order moral insights for the ethics of self-driving cars, but such a solution is difficult to come by. More plausible is the pessimistic solution, and it teaches us a methodological lesson. The lesson is that machine ethicists should discount case intuitions and instead rely on intuitions and judgments at a higher level of generality. (shrink)

The ethical dilemma of whether autonomous vehicles should protect the passengers or pedestrians when harm is unavoidable has been widely researched and debated. Several behavioral scientists have sought public opinion on this issue, based on the premise that EDs are critical to resolve for AV adoption. However, many scholars and industry participants have downplayed the importance of these edge cases. Policy makers also advocate a focus on higher level ethical principles rather than on a specific solution to EDs. But conspicuously (...) absent from this debate is the view of the consumers or potential adopters, who will be instrumental to the success of AVs. The current research investigated this issue both from a theoretical standpoint and through empirical research. The literature on innovation adoption and risk perception suggests that EDs will be heavily weighted by potential adopters of AVs. Two studies conducted with a broad sample of consumers verified this assertion. The results from these studies showed that people associated EDs with the highest risk and considered EDs as the most important issue to address as compared to the other technical, legal and ethical issues facing AVs. As such, EDs need to be addressed to ensure robustness in the design of AVs and to assure consumers of the safety of this promising technology. Some preliminary evidence is provided about interventions to resolve the social dilemma in EDs and about the ethical preferences of prospective early adopters of AVs. (shrink)