A runaway trolley is about to collide into five unaware workers. There is no time to warn them, but you have access to a switch that can derail the trolley onto a track with just one worker. [1]
The question is: Would you use the switch?
This iconic thought-experiment, known as the “Trolley Problem”, is being revived again with self-driving cars.[2] Should a self-driving car save its five passengers from a death fall by running into and killing one pedestrian?
The utilitarian answer of sacrificing one for the many, seems straightforward. Would it however still be for the greater good if the car was full of criminals? Put it differently, would you buy a self-driving car that may potentially kill you for the greater good? From a game theory perspective the Trolley Problem is a zero-sum game and we humans cannot solve it.[3] But, does that really matter?
Some argue that because self-driving cars, unlike humans, do not act by using their intuition, but by executing somebody’s lines of code, we should first settle on the Trolley Problem before we promote autonomous driving.[4]
While this idea may hold theoretically, in reality it does not. In real-life trolley problems are exceptionally rare.[5] They require an inevitable fatality with just one alternative path, which would also cause a fatality. Besides, we are still miles away from having an AI with the capabilities to assess and act upon such scenarios.[6]
We live in a world where 3,287 people die in car accidents every day.[7] Car accidents will happen with or without autonomous vehicles. However, considering that the majority of these accidents are caused by humans [8], we will be a lot better off with self-driving cars. Unlike humans, self-driving cars have the potential to communicate within split-seconds and to “deep learn” from each other in order to avoid these zero-sum scenarios.[9]
Eventually self-driving cars will hack the Trolley Problem for us. Not by making better moral judgements, but by knowing how to avoid the Trolley Problem all together. In a sense, the only true utilitarian act we humans can undertake is to promote autonomous driving.
How do you see the future of autonomous driving? Apart from the Trolley Problem there are still many challenges on the road ahead, such as people losing their jobs and cars being hacked. Feel free to share your ideas in the comment section below.
If you liked the Trolley Problem, watch this episode of Mind Field (from Vsauce) in which they perform the experiment in real-life. Curious to what you would do in a similar situation, check out MIT’s Moral Machine. And if you can’t get enough of these ethical dilemmas, check out these modern Trolley Problem variations.
Sources:
[1]https://www.spaceship.com.au/blog/2017/Trolley-Problem-Ethical-Self-Driving-Cars
[2]https://www.theguardian.com/technology/2016/aug/22/self-driving-cars-moral-dilemmas
[3]http://fairware.cs.umass.edu/papers/Holstein.pdf
[4]http://blogs.cornell.edu/info2040/2017/09/19/the-moral-dilemma-of-self-driving-cars
[5]https://www.weforum.org/agenda/2018/01/why-we-have-the-ethics-of-self-driving-cars-all-wrong/
[6]https://www.washingtonpost.com/news/innovations/wp/2015/12/01/googles-leader-on-self-driving-cars-downplays-the-trolley-problem/?utm_term=.53d8bbfdee9d
[7]http://asirt.org/Initiatives/Informing-Road-Users/Road-Safety-Facts/Road-Crash-Statistics
[8]http://cyberlaw.stanford.edu/blog/2013/12/human-error-cause-vehicle-crashes
[9]https://www.vox.com/2016/6/13/11896166/self-driving-cars-ethics
Very interesting and nice-written blog! It makes me wonder what kind of coding and techniques will be used that able self-driving cars to avoid dilemma situations in which they need to choose between human lives.