Ethics Explainer: The Trolley Problem

Imagine you are the driver of a runaway trolley (tram) that is barrelling down the tracks towards five workers. You have the option to do nothing and let five people be run over, or the option to divert the trolley onto another track and kill one person on the other track instead. 

This is Philippa Foot’s famous trolley problem, popularised by Judith Jarvis Thompson. This thought experiment encourages us to interrogate our moral intuitions and think about the moral differences between actively causing death (pulling a lever to get the trolley to change tracks) and passively or indirectly causing death (doing nothing, allowing the trolley to kill five people). 

The doctrine of double effect

The problem comes from Foot’s most influential paper The Problem of Abortion and the Doctrine of Double Effect, published in 1967. Here, she talks about the Doctrine of the Double Effect, which explains why some actions might be permissible because of their intent and positive outcomes, despite their foreseeable negative outcomes as well. The trolley problem is one example where the doctrine of double effect can be employed, but she also uses various other cases.  

Foot says, “The words ‘double effect’ refer to the two effects that an action may produce: the one aimed at, and the one foreseen but in no way desired. By ‘the doctrine of the double effect’ I mean the thesis that it is sometimes permissible to bring about by oblique intention what one may not directly intend.” 

In this way, it’s arguable that pulling the lever in the trolley problem is permissible because of the intention to prevent the trolley from running over five people, despite the foreseeable (but undesired) result of the action also being that the trolley runs over one person. 

Utilitarianism

This thought experiment has become commonly associated with the ethical framework of consequentialism, or more specifically utilitarianism: a framework that prioritises the greatest good for the greatest number. Philosophers have used the trolley problem, and endless variations of it, to argue both for and against utilitarianism, with classic utilitarian views arguing that it would be morally preferable (and thus obligatory) to pull the lever. 

There has also been much discourse about whether trolley problems are a useful measure of utilitarianism at all. The objections tend to mirror general objections to the emphasis of thought experiments – specifically, that they are unrealistic and unrepresentative and that they tend to elicit different psychological reactions than real moral situations. Despite this, they remain prevalent in philosophy, law, psychology and everyday media. 

 The self-driving trolley problem

One of the main criticisms of the trolley problem is its apparent inapplicability to reality: When in our everyday lives are we faced with such black-and-white situations where the outcomes of our actions are so prescribed and dire? 

Historically, this may have been a compelling argument, but for many years technology has been advancing to such a degree that the trolley problem is quickly becoming a reality. Not because there has been an increase in out-of-control trains, but because of self-driving vehicles. 

If an autonomous car has to “choose” between hitting one person head-on or steering off the road towards a group of people in an unavoidable collision situation, which is preferable? In a regular car, we rely on reflex and deal with the fallout with relatively little moral culpability. Autonomous cars are preprogrammed, though. They can be designed to deal with emergency situations decisively and in a more “thoughtful” way than any human would be able to respond. 

Part of this design essentially entails reckoning with many variations on the trolley problem. Researchers from MIT ran a survey called Moral Machine from 2016–2020 that aimed to gauge public opinion on how machines should make decisions when faced with these kinds of moral dilemmas. The results varied greatly between the 200+ participating countries – with, for example, some cultures prioritising the young over the old, and others showing the opposite preference – demonstrating a clear cultural relativity in how we resolve ethical trade-offs. 

This raises questions about implementation and enforcement. Should there be global standards for autonomous vehicle logic? Should we allow a level of customisation to reflect cultural or personal moral preferences? 

What was once a thought experiment reserved for philosophy classrooms is now a very live debate, one that will only become more important as the technology spreads.