Autonomous Vehicles and the Attribution of Moral Responsibility

Abstract

With the imminent advent of autonomous vehicles (AVs) comes a moral dilemma: How do people assign responsibility in the event of a fatal accident? AVs necessarily create conditions in which “drivers” yield agency to a machine. The current study examines how people make attributions of blame and praise in this context. Varying the features of AV technology affected how responsible a “driver” (who purchased the vehicle) is perceived to be following a deadly crash. The findings provide support for agency and commission as crucial bases of moral judgment. They also raise questions about how morally contradictory actions are perceived and underscore the need for research examining how moral responsibility is distributed among multiple potentially culpable agents. Pragmatically, these findings suggest that regulating (or declining to regulate) how AVs are programmed may strongly influence perceptions of moral and legal culpability.

Risk of Injury in Moral Dilemmas With Autonomous Vehicles

“…This framework acknowledges that some people adjust their behavior according to what others in society do (Crutchfield, 1955;Kollock, 1998;Hornsey et al, 2003;Kundu and Cummins, 2013;Rand and Nowak, 2013;Bostyn and Roets, 2017;Gogoll and Müller, 2017). These social aspects of moral decision-making are absent, or at least only implicit, in common formulations of AV moral dilemmas (Pan and Slater, 2011;Francis et al, 2016;Awad et al, 2018;Faulhaber et al, 2019;McManus and Rutchick, 2019). This formulation, thus, is likely to have higher ecological validity (Conitzer et al, 2017;Gogoll and Müller, 2017).…”

“…To avoid the ethical difficulties of running experiments about life and death situations, most of the earlier work asked participants to make a moral choice or judgment in the context of hypothetical (Greene et al, 2001;Mikhail, 2007) or presumably real (Rutchick et al, 2017;Bostyn et al, 2018) moral dilemma scenarios. In the context of selfdriving cars, prior work has studied participants' choices in hypothetical scenarios (Awad et al, 2018;McManus and Rutchick, 2019) and virtual reality simulations (Pan and Slater, 2011;Francis et al, 2016;Faulhaber et al, 2019). Following this tradition, in this paper, we present several experiments where we manipulate the likelihood of injury to drivers and pedestrians in a moral dilemma scenario involving AVs and measure participants' decision-making.…”

“…Imagine that an AV is driving through a tunnel and is suddenly faced with several pedestrians crossing in the middle of the road; should the AV swerve against the wall—injuring or even killing the driver—or continue moving forward—injuring or even killing the pedestrians? Prior research suggests that in this situation, many people make the choice that saves the most lives and, in this case, possibly results in the driver’s death to be the appropriate choice ( Bonnefon et al, 2016 ; Awad et al, 2018 ; Faulhaber et al, 2019 ; McManus and Rutchick, 2019 ). However, paradoxically, people also report a preference to purchase for themselves an AV that prioritizes the safety of the driver ( Bonnefon et al, 2016 ).…”

“…Prior research has mostly focused on moral dilemmas where the outcomes lead to the death of the targeted human(s) with 100% certainty ( Bonnefon et al, 2016 ; Faulhaber et al, 2019 ; McManus and Rutchick, 2019 ). Whereas these extreme cases are important to understand people’s moral decisions, in reality, many outcomes are unlikely to lead to certain death.…”

Improving the Public's Perception of Autonomous Vehicles by Communicating the Consistency of Autonomous Vehicle Algorithms

“…Finally, in Study 3, we assess how changes in pilot agency (i.e., an AV taking control of a vehicle from a human driver or vice versa) influence judgments of AVs and human drivers performing either deontological or utilitarian actions. Based on past work [17], we hypothesize that AVs would be judged especially harshly when taking control from a human driver.…”