Trolley cases and autonomy violation

Trolley cases are widely considered central to the ethics of autonomous vehicles. We caution against this by identifying four problems. (1) Trolley cases, given technical limitations, rest on assumptions that are in tension with one another. Furthermore, (2) trolley cases illuminate only a limited range of ethical issues insofar as they cohere with a certain design framework. Furthermore, (3) trolley cases seem to demand a moral answer when a political answer is called for. Finally, (4) trolley cases might be epistemically problematic in several ways. To put forward a positive proposal, we illustrate how ethical challenges arise from mundane driving situations. We argue that mundane situations are relevant because of the specificity they require and the scale they exhibit. We then illustrate some of the ethical challenges arising from optimizing for safety, balancing safety with other values such as mobility, and adjusting to incentives of legal frameworks.

Recent years have witnessed a dramatic increase in the scrutiny of artificial intelligence applications. While we applaud this deserved attention, we are concerned that a substantial portion of the ethical theorizing in this domain has proceeded with insufficient attention to the details of the technology involved. This paper argues for a new approach to reasoning about the ethics of machine learning. We make our case by way of an extended evaluation of the literature on autonomous vehicles and trolley cases. We show that theorizing on this topic has relied on a mistaken assumption about how the engineering of autonomous vehicles is likely to proceed. We argue that identifying this mistake allows us to see why trolley cases are not immediately useful in answering moral questions about autonomous vehicles, and it reveals how to proceed when crafting moral arguments about any technology that makes use of machine learning.

The hypothetical scenarios generally known as trolley problems have become widespread in recent moral philosophy. They invariably require an agent to choose one of a strictly limited number of options, all of them bad. Although they don’t always involve trolleys / trams, and are used to make a wide variety of points, what makes it justified to speak of a distinctive “trolley method” is the characteristic assumption that the intuitive reactions that all these artificial situations elicit constitute an appropriate guide to real-life moral reasoning. I dispute this assumption by arguing that trolley cases inevitably constrain the supposed rescuers into behaving in ways that clearly deviate from psychologically healthy, and morally defensible, human behavior. Through this focus on a generally overlooked aspect of trolley theorizing – namely, the highly impoverished role invariably allotted to the would-be rescuer in these scenarios – I aim to challenge the complacent twin assumptions of advocates of the trolley method that this approach to moral reasoning has practical value, and is in any case innocuous. Neither assumption is true.

While it is permissible to switch the trolley in the classic Switch case, it is not permissible to push the stranger in the classic Footbridge (aka, ‘Push’) case. But what may we do in cases that offer both a ‘switch-like’ option and a ‘push-like’ option? Surprisingly, we may choose the push-like option, provided that it has better consequences than the switch-like option. We arrive at this conclusion by taking ourselves seriously – not just as agents who might redirect threats – but as threats who might be redirected by agents.