The social dilemma of autonomous vehicles (1510.03346v2)

Abstract: Autonomous Vehicles (AVs) should reduce traffic accidents, but they will sometimes have to choose between two evils-for example, running over pedestrians or sacrificing itself and its passenger to save them. Defining the algorithms that will help AVs make these moral decisions is a formidable challenge. We found that participants to six MTurk studies approved of utilitarian AVs (that sacrifice their passengers for the greater good), and would like others to buy them, but they would themselves prefer to ride in AVs that protect their passengers at all costs. They would disapprove of enforcing utilitarian AVs, and would be less willing to buy such a regulated AV. Accordingly, regulating for utilitarian algorithms may paradoxically increase casualties by postponing the adoption of a safer technology.

• The paper shows that while participants rate utilitarian AVs as morally superior, they prefer self-protective vehicles for personal safety.
• The paper employs six MTurk studies to uncover a social dilemma where public endorsement of utilitarian algorithms contrasts with personal buying preferences.
• The paper highlights that enforcing utilitarian regulations on AVs may delay market adoption, potentially reducing overall traffic safety improvements.

The Social Dilemma of Autonomous Vehicles

This paper, authored by Jean-François Bonnefon, Azim Shariff, and Iyad Rahwan, explores the ethical complexities associated with the deployment of autonomous vehicles (AVs) and introduces a data-driven approach to understanding public opinion on these technologies. Utilizing six Mechanical Turk (MTurk) studies, the authors investigate the moral preferences of participants when faced with various traffic dilemmas involving AVs. The primary focus is on the utilitarian approach, which aims to minimize casualties, versus a self-protective approach, which prioritizes the safety of the passengers.

Summary of Findings

1. Utilitarian vs. Self-Protective AVs:
   • Participants generally agreed that utilitarian AVs, which would sacrifice their passengers to save a greater number of lives, were more moral.
   • 76% of participants in Study 1 felt it was more moral for AVs to sacrifice one passenger to save ten pedestrians.
   • Despite recognizing the moral superiority of utilitarian AVs, participants expressed a preference for riding in self-protective AVs that would prioritize their safety over that of pedestrians.
2. Social Dilemma:
   • The studies highlight a classic social dilemma: individuals approve of utilitarian AVs for others but prefer self-protective AVs for themselves.
   • Study 3 revealed that participants' willingness to buy an AV significantly decreased if the AV was programmed to sacrifice them and their family members.
3. Perception of Government Regulations:
   • Participants were generally reluctant to accept regulations that would enforce utilitarian algorithms in AVs.
   • In Study 6, participants indicated a significantly lower likelihood of purchasing a regulated AV as opposed to an unregulated one.
4. Impact on AV Adoption:
   • The authors speculate that regulating AV algorithms to enforce a utilitarian approach may paradoxically delay the adoption of AVs, potentially leaving more people at risk due to human error in traditional driving.
   • This presents a significant challenge for policymakers who aim to balance moral integrity, public safety, and market adoption.

Implications and Future Directions

The findings of this paper carry substantial implications for both practical applications and theoretical considerations. Practically, the preference for self-protective AVs highlights a potential obstacle in the widespread adoption of AVs, which could otherwise lead to significant reductions in traffic accidents and increases in traffic efficiency.

Theoretically, the paper underscores the complexity of embedding moral decision-making in machine algorithms. It raises questions about the ethical frameworks that should guide these decisions and the potential need for a collective societal agreement on these frameworks.

Additionally, the research suggests that moral algorithms for AVs will need to contend with more nuanced decisions, including scenarios with uncertain outcomes and considerations of blame assignment. This paves the way for further investigations into how AVs should handle situations involving expected risk and value, and how they might integrate factors such as the age of individuals involved in potential accidents.

Future Research

Future studies should expand the scope to include a more diverse demographic to better understand cultural differences in moral attitudes towards AVs. Moreover, exploring the long-term shifts in public opinion as AVs become more prevalent could provide valuable insights for manufacturers and policymakers. Finally, interdisciplinary collaborations between computer scientists, ethicists, and legal experts will be crucial in formulating comprehensive guidelines and regulations for the ethical deployment of AVs.

Conclusion

This paper significantly contributes to the discussion on the ethical programming of AVs by providing empirical evidence on public moral preferences. It reveals a social dilemma where individuals endorse utilitarian principles in theory but prefer self-protective measures in practice. This discordance presents a substantial challenge for the implementation of AV technology and suggests that achieving a balance between moral principles and public acceptance will be essential for the successful deployment of AVs.