“The Trolley Problem” is an ethical quandary invented by philosopher Philippa
Foot in the 1960s and frequently modified by various audiences since then. The original quandary is this: Imagine you’re driving a runaway train that
is about to collide with a set number of people on the tracks. You could switch tracks and kill a smaller
number of people on alternative tracks.
Do you switch tracks? The
question can be modified to determine how much you are influenced by various
variables such as relationship with the individuals, worth of those
individuals, and the sense of agency you feel in the resulting deaths. If you use a truly utilitarian approach to making
the decision, you should always choose the track that would kill the fewest
people (or at least the people with the lowest societal value). However, the results of this thought
experiment have repeatedly shown that people are not making purely utilitarian
decisions. In the last few decades,
neuroscientists have begun to determine which brain areas are active while
people are making these decisions, in an effort to better understand brain areas
involved in complex decision-making.
This new “scientific” dimension of the question has led to a resurgence
of interest in the trolley problem.
Although the trolley problem has
often been critiqued as being an unrealistic situation, programmers working on
self-driving cars must now wrestle with very similar ethics. Programmers have the advantage of plenty of
time to contemplate the best course of action in various situations, unlike
drivers who must react in the moment. If
a self-driving car suddenly detects a small child that runs in front of it,
should the car swerve, despite knowing it would hit the brick wall and kill its
driver? Public opinion surveys suggest
that people want self-driving cars to prioritize the greatest number of lives,
but they’d be much more likely to buy a self-driving car if their particular
self-driving car prioritized the life of the driver. Artificial intelligence will allow us to make
the “best” decisions more often; the problem is that as humans, even with
plenty of time to think about those decisions, we still struggle to know what
those best decisions are. - SLB
Recommended
sources for further engagement:
Moral Machine. by the Scalable Cooperation at
MIT Media Lab. 2016.
Simulation that allows you to decide between two options for what a
self-driving car should do, and then compares your responses to those who have
answered previously
Can You Program Ethics Into a Self-Driving Car? by Noah J. Goodall in IEEE
Spectrum. 2016.
Examines the ethics of self-driving cars
The Social Dilemma of Autonomous Vehicles. by Jean-François Bonnefon, Azim Shariff, and Iyad Rahwan in Science. 2016.
The peer reviewed article that asked people what they would want cars to do in general and what they would want their particular car to do (but you need access to Science to read it)
The Trolley Problem and the Evolution of War. by Peter
Reiner at the Neuroethics at the Core blog.
2011.
Briefly describes the trolley problem in general, and then applies the model
to decisions made in war, including the use of drones which is another AI
application
The Rational Vulcan. by Dave Johnson at the Neuropoly
blog. 2010.
Some history, other ethical thought problems, and
a description of brain areas involved when answering the trolley problem
The Hypocrisy of Professional Ethicists. by Emma Green
in The Atlantic. 2015.
Summarizes findings that suggest that studying ethics (and thus engaging in
thought experiences like the trolley problem) does not result in making more
ethical decisions
