Self-driving cars are rapidly becoming a reality. In a report by the Economist in August, it was speculated that self-driving cars could reduce the number of vehicles on the road by up to 90 per cent. A study by the Eno Center for Transportation estimated that if 90 per cent of cars on the road were autonomous the number of accidents per year would reduce by 75 per cent. However, autonomous vehicles bring to mind a classic thought experiment from the field of ethics: the Trolley problem.
Should a self-driving car kill its passengers in order to save 10 people? What should a car be programmed to do in the case of unavoidable loss of life?
Imagine three scenarios that involve unavoidable harm. A car can swerve and kill one person instead of 10. A car can swerve and kill its occupant or stay on course and kill one person. A car can swerve and kill its occupants or stay on course and kill 10 pedestrians. What should a car do in these situations? The simplest and most utilitarian solution is to minimize the number of casualties even if it means the death of the passengers. In other words, a car must kill its five occupants if it means that the deaths of more than five people can be prevented. Who then, would buy a self-driving car if it was programmed to kill them in the case of emergencies?
A study conducted in October examined people’s opinions on this utilitarian approach. In general, people were comfortable with autonomous vehicles minimizing the loss of life. However, they liked the idea of others driving these utilitarian cars more than being in one themselves.
The main problem with autonomous cars is not whether they must kill its passengers, but rather who is to blame in the case of an accident. Can the accident be blamed on the driver if it was the result of a bug in the software? Or is it the manufacturer’s fault for not making its system completely fool proof?
The most important fact that the public must understand about self-driving cars is that they are exponentially more aware of their surroundings than humans. Self-driving cars aren’t distracted by conversations, don’t zone out, and do not feel drowsy. Self-driving cars today are equipped with radar and/or LIDAR, which are remote sensing technologies that enables a computer to “see”. These cars can not only see perfectly during the day, but also at night. They can see past partially solid materials, obstacles, and extract information from objects like road signs and flashing lights. In other words, a self-driving car is too keenly aware of its surroundings to find itself in situations where death is unavoidable.
In addition, the next step in the autonomous vehicle technology is data sharing. The V2X communication system (vehicle to X), would allow vehicles to share their location, course information, and data from their sensors with other cars, or any other infrastructure system. This will allow the software to visualize a model of the world around them in excruciating detail. When a significant percentage of all vehicles are equipped with V2X systems, vehicle safety will improve tremendously and the world will be at the brink of fully autonomous driving.
Even today, features like cruise control, automatic braking and lane keeping are used by many automotive manufacturers. Autonomous vehicles will not only improve road safety and boost productivity but also improve general quality of life. However, they bring with them their own set of problems and ethical dilemmas. The question that remains is not whether the world will adopt self-driving cars, but rather when and how.