Autonomous Driving Ethics Dilemma Database EthicsDrive Deep Dive: 100,000 Real Scenarios Reveal Human Moral Choice Patterns

Autonomous Driving Ethics Dilemma Database EthicsDrive Deep Dive: 100,000 Real Scenarios Reveal Human Moral Choice Patterns

When an autonomous vehicle faces an unavoidable collision, should it prioritize protecting its passengers or pedestrians? Should it swerve toward one elderly person or three young people? These "trolley problems" have been at the center of autonomous driving ethics discussions for the past decade, but most have remained at the level of thought experiments.

EthicsDrive, the upgraded version of MIT's Moral Machine project, is pushing these discussions into the realm of empirical evidence. The database contains 100,000 ethical dilemma scenarios extracted from real traffic accident records around the world — each with detailed road conditions, participant information, and actual human driver decision records.

Through large-scale surveys of 2.5 million respondents across 42 countries, EthicsDrive revealed significant cultural differences. East Asian respondents were more inclined to choose options that "protect the majority" (even at the expense of minorities), while European respondents adhered more strongly to the principle of "not actively harming anyone." Latin American respondents placed greater emphasis on age as a factor in decision-making (prioritizing the protection of young people), while Middle Eastern respondents placed greater emphasis on gender.

"This data is not meant to tell autonomous vehicles what to do," said MIT Professor Iyad Rahwan. "It is meant to reveal the complexity of human moral intuition — any attempt to program autonomous driving ethics with a single simple rule is an oversimplification."

EthicsDrive's data has been incorporated by the European Commission as a reference for developing autonomous driving safety regulations. The EU's new rules require autonomous driving systems to be able to explain their decision-making logic in ethical dilemmas, but do not prescribe specific moral choice standards.

However, researchers caution that "following the crowd" does not equal "being correct" — if the majority of people make a certain moral choice in a given scenario, it does not mean that choice is morally justified. Autonomous driving ethics programming cannot simply replicate human moral intuition.