An ethical decision-making framework for autonomous vehicles based on public moral preferences

Area of Science:

• AI Ethics and Autonomous Systems
• Human-Computer Interaction
• Transportation Safety

Background:

• Autonomous vehicle (AV) ethical decision-making is crucial for public acceptance and technology adoption.
• Current AV algorithms lack integration of public opinion in unavoidable ethical dilemmas.
• Public trust in AVs is contingent on their perceived ethical rationality.

Purpose of the Study:

• To propose an ethical decision-making algorithm framework for autonomous vehicles grounded in public moral preferences.
• To model societal moral consensus into machine-executable language for AVs.
• To enhance the ethical alignment of AVs with human values and improve safety.

Main Methods:

• Identified five core moral preferences: reducing overall injury, reducing individual injury, ensuring equal care, protecting the compliant, and avoiding active harm.
• Conducted a large-scale online survey to quantify road user injury severity and model moral preferences.
• Developed and validated an ethical decision-making algorithm using 4,860 diverse traffic scenarios.

Main Results:

• The proposed algorithm closely aligned AV decisions with identified public moral preferences.
• Demonstrated significant reductions in average individual injury severity (52%), injury severity standard deviation (15%), and injury severity for compliers (20%) and non-occupiers (24%) compared to a baseline.
• The framework showed high scalability and alignment with societal ethical considerations.

Conclusions:

• Integrating public moral views into AV ethical algorithms is feasible and effective.
• The developed framework offers a scalable solution to improve AV ethical decision-making and public acceptance.
• This research provides a foundation for culturally adaptable AV ethical frameworks, facilitating technology promotion.