A runaway streetcar is hurtling towards five unsuspecting workers. Do you pull a switch to divert the trolley onto another track, where only one man works alone? Or do you do nothing?
This haunting choice is a variation of the “trolley problem,” an iconic philosophical thought experiment. (If you’ve never heard of it before, try this.) Puzzling, ridiculous, and oddly irresistible, this imaginary scenario has profoundly shaped our understanding of right and wrong. In the past 40 years it has occupied the attention of brilliant minds, from academic ethicists to moral psychologists to engineers. It has helped them try to answer profound questions—how do we act, and how should we? But in its fifth decade, is the trolley problem starting to show its age?
Thought experiments have been essential in scholarly discourse since the ancient Greeks and Romans. Most remain secluded in the ivory tower, while a handful, like the simultaneously-dead-and-alive Schrödinger’s cat or the Prisoner’s Dilemma, escape the realm of the academic to become powerful cultural mainstays—not always for well-understood reasons.
Curiously, the first thinkers to popularize and analyze the trolley problem were women—in general, rare voices in philosophy. Its story begins in 1967 at Oxford University, when the “grand dame of philosophy” Philippa Foot devised the example of the runaway streetcar—“tram” in England, “trolley” in the U.S.—while discussing the permissibility of abortion.
A few years later, Judith Thomson, a philosopher at the Massachusetts Institute of Technology,  coined the term “trolley problem” and created what would become its two most famous variants, the “footbridge” and the “switch.” In the “footbridge” scenario (also known as “fat man”), the streetcar is heading towards five workers, but this time you’re on a footbridge over the track. Standing precariously close to the edge of the bridge next to you is a very large man, who, if he happened to topple onto the track below, could stop the trolley before it reaches the five. Do you push him? Thomson’s writing sparked so much interest in the philosophical community that a sub-discipline of “trolleyology” emerged during the ‘70s and ‘80s.
The trolley dilemmas vividly distilled the distinction between two different concepts of morality: that we should choose the action with the best overall consequences (in philosophy-speak, utilitarianism is the most well-known example of this), like only one person dying instead of five, and the idea that we should always adhere to strict duties, like “never kill a human being.” The subtle differences between the scenarios provided helped to articulate influential concepts, like the distinction between actively killing someone versus passively letting them die, that continue to inform contemporary debates in law and public policy. The trolley problem has also been, and continues to be, a compelling teaching tool within philosophy.
By the late ‘90s, trolley problems had fallen out of fashion. Many philosophers questioned the value of the conclusions reached by analyzing a situation so bizarre and specific.
It wasn’t clear trolleys could ever find a life out of the pages of academic journals until one philosophy graduate student, Joshua Greene, revived them with the modern techniques of neuroscience.
Greene decided to slide people into an fMRI machine to glimpse what happened in their brains when faced with different trolley-problem scenarios. He ultimately found that the answers people gave correlated with how emotionally engaged they felt with the dilemma. The decision to pull the switch was related to activity in the prefrontal cortex (associated with cool, conscious deliberation), while the decision not to push the fat man involved areas like the amygdala, associated with strong emotional reactivity.
The paper that reported these results in 2001 inspired hundreds more studies using the trolley problem to study moral decision-making—and also piqued interest from people in other fields, like sociologists, economists, and anthropologists. Greene and others have used trolleys to conclude that strong mental imagery and visceral emotions make us more likely to make an intuitive decision (“never kill a human being”), as opposed to a more mathematical calculated decision (five lives versus one). Greene’s research also suggests that the reason people are less likely to push the fat man than to flip a switch is because we all possess a biologically pre-programmed emotional aversion to delivering harm personally (touching a man as you shove him off a bridge) as opposed to impersonally (yanking a lever).
The ethicist Peter Singer cites Greene's research to support some of his positions about why we ought to make greater sacrifices for problems that may seem distant, like world poverty or a disease raging halfway around the globe. Singer argues that we shouldn’t avoid our moral obligations to someone just because they live too far away to engage our brain’s emotional machinery.
Despite the insights Singer and others gained from the trolley problem, many psychologists, like philosophers before them, eventually began to tire of it. The University of California, Irvine, psychologist Christopher Bauman and his colleagues summarized the problem in a paper last year: Researchers have noted that trolley-problem scenarios frequently cause study participants to laugh, meaning they aren't taking the experiment seriously—possibly because the scenarios don't mirror believable, real-life moral dilemmas. Most of us probably won’t find ourselves in a strangely bare landscape, coincidentally placed next to a life-giving switch, or forced to decide whether or not to push a man off a bridge.
But recently, trolley problems have found new life in a more realistic application: research on driverless cars.
Chris Gerdes, a professor of mechanical engineering at Stanford and the director of their automotive-research center, has spent years on algorithms for automated vehicles, figuring out how these cars should handle emergencies and make decisions acceptable to society. When I spoke to Gerdes, he had just returned from a test drive on the California streets. He explained that many of the situations driverless cars will face involve conflicting priorities. When a vehicle has no option but to have a collision, which collision is it going to have? This is where trolleys come in.
“These problems we were trying to solve were not simply technical issues, but things that you could actually turn to philosophy for some insight,” Gerdes told me. “In those cases where loss of life may be inevitable—and there will be situations like that—we want the car to make a reasonable decision.” Gerdes also thinks the trolley problem is a useful springboard: “[It] is one way of highlighting the fact that you eventually reach a point where you have to make some decisions, and not everybody will agree.”
Gerdes has been working with a philosophy professor, Patrick Lin, to make ethical thinking a key part of his team’s design process. Lin, who teaches at Cal Poly, spent a year working in Gerdes’s lab and has given talks to Google, Tesla, and others about the ethics of automating cars. The trolley problem is usually one of the first examples he uses to show that not all questions can be solved simply through developing more sophisticated engineering. “Not a lot of engineers appreciate or grasp the problem of programming a car ethically, as opposed to programming it to strictly obey the law,” Lin said.
But the trolley problem can be a double-edged sword, Lin says. On the one hand, it’s a great entry point and teaching tool for engineers with no background in ethics. On the other hand, its prevalence, whimsical tone, and iconic status can shield you from considering a wider range of dilemmas and ethical considerations. Lin has found that delivering the trolley problem in its original form—streetcar hurtling towards workers in a strangely bare landscape—can be counterproductive, so he often re-formulates it in terms of autonomous cars:
You’re driving an autonomous car in manual mode—you’re inattentive and suddenly are heading towards five people at a farmer’s market. Your car senses this incoming collision, and has to decide how to react. If the only option is to jerk to the right, and hit one person instead of remaining on its course towards the five, what should it do?
It may be fortuitous that the trolley problem has trickled into the world of driverless cars: It illuminates some of the profound ethical—and legal—challenges we will face ahead with robots. As human agents are replaced by robotic ones, many of our decisions will cease to be in-the-moment, knee-jerk reactions. Instead, we will have the ability to premeditate different options as we program how our machines will act. For philosophers like Lin, this is the perfect example of where theory collides with the real world—and thought experiments like the trolley problem, though they may be abstract or outdated, can help us to rigorously think through scenarios before they happen. Lin and Gerdes hosted a conference about ethics and self-driving cars last month, and hope the resulting discussions will spread out to other companies and labs developing these technologies.
But could trolley problems, beyond helping us to design technology, also serve as a tool for everyday self-improvement? The philosopher and psychologist Eric Schwitzgebel, who has studied the behavior of ethics professors, found that philosophical expertise does little to change their moral behavior—for example, they’re no more likely than others of similar social background to donate to charity or stop eating meat. Schwitzgebel doubts that spending time puzzling over trolleys can actually help a person make better moral decisions. But he still thinks it’s useful to hold onto thought exercises like the trolley for research purposes, even if they don’t really seem to change the way people behave in the real world, or are an imperfect analogue to the messy decisions they typically face.
“We as psychologists and experimental philosophers should be pretty careful about how subjects are interpreting [the trolley problem],” Schwitzgebel told me, “and there is a certain lack of external validity to it. On the other hand, it’s a nice, clean problem.” This simplicity, Schwitzgebel believes, is what makes it such an incisive tool for scientific investigation. Scientists still need the trolley problem, in addition to more realistic scenarios, he said, because “basic research into morality is an important part of the human condition.”
It’s hard to tell if this means the trolley problem will make another resurgence, though Judith Thomson herself wrote to me in an email: “I don’t for a moment think the trolley problem is approaching its end.”
As for the rest of us, who may be weary from wondering why we couldn’t just warn the workers to get out of the way and avoid the whole mess in the first place, my advice is this: Just keep sipping your latte.