Virtual Morality: Using Virtual Reality to Study Moral Behavior in Extreme Accident Situations

Virtual Reality (VR) technologies are widely employed to investigate human behavior in dangerous situations that cannot be safely reproduced in the real world, allowing researchers to study in an ecological way complex scenarios such as training for risky jobs, safety procedures, emergencies and, more recently, moral dilemmas in driving context. Understanding how people act when facing severe accidents involving unavoidable collisions has extremely important implications for the design and development of the "decisional system" of Autonomous Vehicles (AVs). However, previous studies have not focused on the differences between being the driver acting in a complex moral situation or being in a self-driving car that chooses for you. In the present paper, we described a case study that uses a first-person virtual reality simulation to investigate people's emotional reactions, perceived sense of responsibility, and acceptability of moral behavior in human and autonomous driving modalities. The main findings showed that participants experienced a high sense of presence in our simulation and react differently to the two driving conditions, showing a greater arousal, a more negative valence, and an increased sense of responsibility when faced moral dilemmas as drivers. Instead, in scenarios that did not involve killing someone (non-moral dilemmas), being in a fully autonomous vehicle was judged less pleasant than being the actual driver. These results suggest that people prefer to be in control only in common driving situations and not when their actions have deadly consequences on other people, suggesting the need to consider emotional factors in studying decision-making applied to autonomous vehicles, as a mean to reach a more complete understanding of people's reactions to this new technology, and to possibly gain insights for the design of autonomous driving systems and, more generally, AI-driven machines.