Moral Luck: Why the Same Act Is Judged by How It Turned Out

Two people can do exactly the same thing, with the same knowledge and the same disregard for the risk, and end up in different moral positions because of something neither controlled. The driver who gets home has done everything the driver who hits someone did. What separates them is whether a pedestrian happened to step off the curb. Most people nonetheless blame the one who hit somebody more, at least when that is the only case they see, and that gap is what philosophers call moral luck.

Bernard Williams introduced the term and Thomas Nagel replied to him; the two essays were reprinted in Nagel's Mortal Questions in 1979 and in Williams's Moral Luck in 1981. Nagel sorted luck into four kinds: luck in how an action turns out, luck in the circumstances a person faces, luck in the character they start with, and luck in the causes that produced them. The demo above is about the first kind, which later writers call resultant luck, because it is the kind an experiment can isolate: hold the person and the act fixed and vary only the result.

The psychological work suggests the gap is not simply a failure of logic. People who read only the version in which someone was hurt infer a riskier drive and a more careless driver than those who read the version in which nothing happened, though the two descriptions give the same facts. Kneer and Machery found that the effect of the outcome on all four of their measures was mediated by how negligent people judged the agent to be, and traced that in turn to differences in the probability people read into the case. The outcome works as evidence about the act, and from one case alone it is hard to see that the evidence was planted by the ending.

Judgments also come apart depending on what is being asked. Cushman found that questions about whether an action was wrong or permissible turn mainly on what the agent believed and wanted, while blame and punishment turn on the causal link to the harm as well. Kneer and Machery found the same split: shown both cases together, most people gave the two drivers the same blame and far fewer gave them the same punishment.

The empirical question is whether ordinary people really do judge the lucky and the unlucky agent differently, and Markus Kneer and Edouard Machery tested it in a paper published in Cognition in 2019. They used pairs of cases of the kind philosophers argue about, including a driver who has been drinking, and asked four questions: whether the action was wrong, how much blame the agent deserved, whether the action was permissible, and what punishment was appropriate. Some participants saw one case only, the design the demo above copies; others saw both cases together.

The two designs gave different answers. Between subjects, with each person reading one version, the outcome moved all four judgments. Within subjects, with both versions in front of the same reader, the asymmetry largely disappeared: in the authors' words, the vast majority judged the lucky and the unlucky agent equally blameworthy, and their actions equally wrong and equally permissible. Punishment was the exception, and was significantly more outcome-dependent than the other three. The authors concluded that the puzzle of moral luck, as a claim about what people reflectively believe, does not arise, and that the between-subjects effect is explained by participants reading different probabilities and different degrees of negligence into the two cases.

Fiery Cushman's 2008 paper in the same journal takes the question apart from another direction. Across a series of experiments he varied four things independently: what an agent desired, what the agent believed, whether the agent caused the harm, and whether the harm occurred. Judgments of wrongness and permissibility followed the agent's mental states; judgments of blame and punishment followed the mental states and the causal connection to the harm as well.

Neither paper settles whether outcome-sensitivity is a mistake. Grading punishment by consequences is written into the law of most countries, and the arguments for it are about deterrence and the public role of punishment rather than about the driver's state of mind. What the experiments show is narrower: the size of the gap depends on whether you can see the other case.

Outcome luck shows up wherever a decision is reviewed after its result is known. The rows below are illustrations, not findings from the studies cited on this page.

The counters all come down to getting the other case in front of you.

• Put the two cases side by side. In Kneer and Machery's experiments the asymmetry mostly vanished when people could see both versions, so writing out the version in which the luck went the other way is the single most effective step.
• Write down the risk before the result is in. A decision log that records what was known, what the odds looked like and what the alternatives were gives you something to review that the outcome cannot rewrite.
• Ask what you would say if the ending flipped. If a near miss would have gone unmentioned, the act was being judged by the ending rather than by the act.
• Separate blame from consequences. You can hold that two reckless drivers were equally wrong and still think the one who caused an injury owes more; saying so out loud keeps the two questions from merging.
• Watch the probability you are reading in. Check whether your estimate of the risk came from the facts of the case or from knowing how it ended.

Moral luck sits next to several biases about outcomes, and the differences are in what is being judged.

• It is not outcome bias, which is about the rated quality of a decision rather than about blame or desert. Outcome bias says a physician's choice looks worse when the patient dies; moral luck says the drunk driver who hurt someone is a worse person. They are measured with different questions and often move together.
• It is not hindsight bias, the sense once an outcome is known that it was predictable all along. Hindsight bias is one of the routes into moral luck, since an agent who should have seen it coming looks more negligent, but it also distorts memory where no blame is at stake.
• It is not the just-world hypothesis, the belief that people get what they deserve. That belief runs from the outcome to the person's character in general; moral luck is the narrower fact that one act gets judged by its result.