Use thought experiments as intuition pumps
Should a self-driving car swerve and kill one pedestrian on the sidewalk to save the five pedestrians on the crosswalk? We could leave this decision to the programmers of the car’s AI. But this seems arbitrary, since the programmer has no particular expertise in ethics. Alternatively, we can train the cars to do exactly what humans do. But humans react fast enough, or just make instinctive desperate acts like slamming the brakes or swerving as they close their eyes and hope for the best. Emulating this behavior wastes the opportunity to program a more considered principle into the car, ahead of time.
What do ethicists say? Two competing ethical theories give different answers to this dilemma. Consequentialism (or utilitarianism), associated with English philosopher Jeremy Bentham (1748-1832), suggests the car should swerve, since the better consequences--fewer deaths--justify the unintended side-effect of killing one person. This contrasts with deontological (rule-based) ethics, commonly associated with German philosopher Immanuel Kant (1724-1804). Swerving could violate a universal rule such as ‘one must not intentionally take any action to end another person’s life.’
What do normal people think? After all, these people will have to support and live under whatever policy we adopt. In 2015, together with my collaborators Jean-François Bonnefon and Azim Shariff, we ran a survey asking people what they think the car should do in this situation. The overwhelming response to our survey was that the car should swerve. Most people, at least in America where we ran the study, are utilitarian.
It’s easy to dismiss this problem. Indeed this was a common response to our studies in this area. Some people just said “why can’t the car just brake?” Of course, if that’s what the car can do, that’s what the car should do.
The driverless car dilemma is what philosophers call a ‘thought experiment.’ It allows us to imagine a scenario (and somewhat suspend disbelief in some of the unrealistic details) in order to focus on the key features of a situation. Thought experiments are useful, since they help us generate and test our intuitions—they are what philosopher Daniel Denett calls ‘intuition pumps.’ They also help us understand a great deal about human moral reasoning. For example, psychologists can place people in brain scanners and present them with thought experiments, to see which parts of the brain get engaged when facing different moral tradeoffs.
References
Dennett, D. C. Intuition Pumps And Other Tools for Thinking. (W. W. Norton & Company, 2013).
Greene, J. Moral Tribes: Emotion, Reason and the Gap Between Us and Them. (Atlantic Books, 2013).