Should we forcibly harvest a man’s organs to save five people? Should we plug the whole of the humanity into a machine that could ensure the maximal amount of aggregate pleasure if we could? Are we obligated to save a child drowning in a pond? These are all famous thought experiments that philosophers have posited over the past century to sway us over to their way of thinking and appeal to our most basic intuitions. But this practice, argues Edouard Machery, is dangerously flawed.
You’ve probably heard of the famous Trolley case. Here is one version: suppose a runaway trolley is about to hit five workers who, by accident, happen to be working on the track. The only way to prevent their death is by pushing a switch that will redirect the trolley onto another track. Unfortunately, there is another worker on this sidetrack who will be killed if the switch is pushed. The question: Is it permissible to push the switch, saving five people but killing one?
You probably know that this scenario, and many other similar ones, have been used to study moral judgment, or as many say, moral intuition both in neuroscience and psychology. You might even know that they have been used to study the design of automated vehicles. But what you perhaps do not know is that Trolley cases have been a mainstay of moral philosophy for decades. They are not only used to study how people think about the right and the wrong or about the permissible and the forbidden, but also about the right and the wrong or about the permissible and the forbidden themselves. The goal is to determine what is morally right and wrong, required, permissible, and forbidden based on these very cases.
Moral philosophy is not an outlier in philosophy in its use of thought experiments such as the Trolley cases. Metaphysicians use them and epistemologists are addicted to them in their effort to understand knowledge, justification, and other important epistemic notions. Here is a famous one, called the “Gettier case”: Suppose you see a clock that says it is 3:00pm and that you come to believe it is 3:00pm as a result. It is in fact 3:00pm, and your belief is thus true, but the clock is broken and always says 3:00pm. The question is: Do you know that it is 3:00pm?
As you would expect from philosophers, there is an intricate and at times heated debate about the exact role of thought experiments and of the judgments or intuitions they elicit in philosophy. Some deny that intuitions play any role in philosophy whilst others claim that philosophers merely argue for their claims about thought experiments (e.g., “Yes, it is permissible to push the switch!”).
These lively debates notwithstanding, most philosophers would agree, I believe, that thought experiments are often used in philosophy and that the judgments philosophers make in response to them are used in some fashion to support or, more commonly, undermine philosophical theories about knowledge, wisdom, fairness, causation, or permissibility.
How people respond to the different versions of Trolley cases depends on the order in which they read them
The tendency of philosophers to rely on thought experiments, however, I believe to be extremely problematic. Twenty years of experimental philosophy—the use of empirical methods, from corpus studies to experiments, to provide evidence relevant to philosophical questions, often using the methods of social psychology—have taught us two things about thought experiments. First, often people respond to them differently across demographic groups, particularly different cultures, and second; small, irrelevant changes in how thought experiments are worded can change entirely how we respond to them.
The extent of demographic and cultural variation in thought experiments is a matter of intense on-going debate. Some experimental philosophers have recently argued that there is in fact surprisingly little demographic variation in how people react to thought experiments. But I demur. Reviewing 100 studies by more than 20 different researchers comprising a total sample size of over 40 million participants, Steve Stich and I have shown that variation is extensive and often very surprising. To give only a few examples: East Asians appear to have different intuitions about proper names from Westerners and their assessment of free will and responsibility differ from the one found in other parts of the world. Women have different intuitions about moral dilemmas such as the Trolley cases from men. Judgments about the intentional nature of side effects seem to differ in some small-scale societies compared to elsewhere in the world. Gen-Xers are more likely to find it permissible to kill someone to save five people than Baby Boomers. And people disagree across cultures about whether aesthetic judgments are true or merely express preferences.
Trivial changes in thought experiments can also move people’s judgments around. This is what psychologists call “framing effects.” Philosophers themselves such as Bernard Williams, Tamar Gendler, and John Hawthorne had noticed this phenomenon, but it is really experimental philosophers who have demonstrated its reality. For instance, how people respond to the different versions of Trolley cases depends on the order in which they read them.
Demographic variation and sensitivity to trivial changes show that, as a whole, judgments about thought experiments are unreliable. Some judgments appear to be stable across demographic groups as well as hard to frame, but in the absence of compelling evidence that a thought experiment elicits this kind of judgment, one should refrain from using thought experiments in philosophical theorizing. Of course, they can still be extremely useful when psychologists or neuroscientists study our psychology, but their use in philosophy should be severely curtailed.
Philosophers have balked at this suggestion perhaps because they cannot imagine doing philosophy in other ways. There is however much philosophy left that does not require appealing to thought experiments. For instance, thought experiments played little role in the important work on causation in philosophy of science for the last twenty years, and non-ideal theories in ethics have no truck with thought experiments.
Thought experiments can still be extremely useful when psychologists or neuroscientists study our psychology, but their use in philosophy should be severely curtailed
Of particular interest is the recent emphasis on conceptual engineering, i.e. on attempts to reform philosophically significant concepts. Concepts may need to be reformed to make them more useful in some areas of inquiry, to avoid mistakes and fallacies, or to remove morally or politically dubious associations. The prospects, challenges, and limitations of this philosophical practice are still being discussed, but thought experiments are not consequential in this approach to philosophy. Rather, conceptual engineering is hospitable to the use of empirical methods to study the concepts that need to be reformed. Reforming a concept often requires a good understanding of that concept.
It is easy to be misled by thought experiments. The judgments they elicit might just be cognitive illusions that would lead thinking astray or they could simply reinforce our parochialist views about philosophical problems. And there is no reason for philosophers to be bound to this old, but definitely not venerable, method for doing philosophy.