Traffic Morals and Autonomous Vehicles
By Graeme Pattison
The ethics of self-driving vehicles deciding who is to live and who is to die in traffic crashes continues to be hotly debated. Of course, there is no simple answer but to assist the discussion a system has been built to generate moral dilemmas and analyse the decisions people make between two adverse outcomes. It is known as the "Moral Machine" although it is actually an online experimental platform developed at the Massachusetts Institute of Technology.
The dilemmas are often variations of "The Trolley Problem" (https://en.wikipedia.org/wiki/Trolley_problem). The machine has gathered 40 million decisions from millions of people in many countries and has given insight into global preferences, individual variations related to demographics, cross-cultural ethical variations and modern institutions and cultural traits (https://www.media.mit.edu/publications/the-moral-machine-experiment/).
You can actually run the experiment yourself using autonomous vehicle situations at this website: http://moralmachine.mit.edu/. Results and conclusions have been published in the Nature Journal (https://www.nature.com/articles/s41586-018-0637-6).
Image from www.technologyreview.com (MIT)
While Level Five fully autonomous vehicles for all road conditions are some time in our future, there is increasing opinion that geofencing to suitable sections of the road network will allow fully AV operation in the near future. I expect that the safety performance of near future AV will exceed that of many human drivers and a tipping point will be reached. From that stage crash drivers may be questioned in court to justify why they selected manual control when AV offered safer outcomes.
A serious concern in my opinion is setting the safety standards and performance of AV. Unfortunately, the logic and programming in these AV safety systems is generally proprietary and kept secret, preventing independent scrutiny. We should as a community insist on standards with independent testing and have at least partial access to the safety algorithms that our lives (and the lives of other road users) will depend on.