Why a harmful side effect looks intentional and a helpful one does not
The idea
Imagine a company chairman who is told that a new programme will raise profits and also harm the environment. He replies that he does not care about the environment, only about profit. The programme goes ahead and the environment is harmed. Did he harm it intentionally?
Now change one word. The programme will help the environment, the chairman says he does not care, it goes ahead, and the environment is helped. Did he help it intentionally?
In 2003 the philosopher Joshua Knobe put these two stories to 78 people in a public park in Manhattan, each person reading one version chosen at random. In the harm version, 82 percent said the chairman acted intentionally. In the help version, 77 percent said he did not, leaving about 23 percent who said he did. The chairman’s knowledge, motive and indifference were identical; only the moral character of the side effect changed. The finding, published in the journal Analysis, became known as the Knobe effectside-effect effect The finding that people judge a foreseen side effect as more intentional when it is harmful than when it is helpful. or the side-effect effect, and it helped launch experimental philosophyexperimental philosophy The use of surveys and experiments to study how people respond to questions that philosophers argue about., the practice of testing philosophers’ claims about ordinary concepts with surveys and experiments.
The result matters because many accounts of intentional action, in philosophy and in law, treat intention as a fact about the mind: what the person wanted, believed or foresaw. If ordinary judgements of intent shift with whether the outcome was good or bad, then either those judgements are biased, or the everyday concept of intention contains moral content that the theories leave out.
Why it might change a decision
The effect predicts a specific, checkable pattern in how people assess decisions with foreseen side effects: incident reviews, performance appraisals, jury deliberations, news coverage of corporate choices. When the side effect is harmful, people are more likely to describe it as done on purpose, even if the decision maker’s stated aim and knowledge would be the same in a helpful case.
One way to test this on any particular judgement is to rewrite the case with the side effect flipped in valence and keep everything else fixed. If the verdict on intent changes, the judgement is tracking the outcome’s moral character as well as the person’s mental state. Whether that is an error depends on which explanation below is right, which is why the test describes the judgement rather than settling it.
How solid is this?
The original evidence was a single small study: one between-subjectsbetween-subjects A study design in which different groups of people each take part in only one of the conditions being compared. survey of 78 people, in one city, using a short written vignettevignette A short written story given to participants in a study, who are then asked to make a judgement about it.. The difference was large and statistically clear, but a single sample of passers-by says little on its own about how general the pattern is.
The effect has since been retested. The X-Phi Replicability Project, led by Florian Cova and colleagues and published in Review of Philosophy and Psychology in 2021, reran 40 experimental philosophy studies from 2003 to 2015, choosing the most cited paper of each year plus two chosen at random. Knobe’s 2003 paper was the most cited of its year, and the project classified its replicationreplication Running a study again, independently, to see whether the original result holds up. as successful. Across all 40 studies the project estimated a replication rate of roughly 70 to 78 percent depending on the criterion used, which the authors describe as relatively high. That project does not settle the cause of the effect, and the vignette format means it measures what people say about stories, not how they judge real events in which they have a stake.
The live dispute is over explanation, and there are three main candidates:
- Moral content in the concept. On this reading, judging whether something was done intentionally partly depends on evaluating it, so the asymmetry reflects how the concept works rather than a mistake.
- Rational inference from norms. Kevin Uttich and Tania Lombrozo, writing in Cognition in 2010, argued that behaviour that breaks a norm, moral or otherwise, tells an observer more about someone’s state of mind than behaviour that conforms. Someone who proceeds despite a norm against harming reveals what they were prepared to bring about; someone whose side effect happens to be good reveals little. On this view the asymmetry is a sensible use of theory of mindtheory of mind The everyday ability to work out what other people believe, want and intend., not a distortion.
- Blame expressed through wording. A further proposal is that people use “intentionally” partly to signal that blame is due, so the answers reflect what speakers want to communicate rather than a separate judgement about the chairman’s mind.
These accounts make different predictions, for example about non-moral norms and about questions that avoid the word “intentionally”, and studies have tested several of them. None has displaced the others. Popular summaries that present the effect as proof that “morality corrupts our reasoning” go beyond the data: the asymmetry is well established, but whether it reflects bias, sound inference or the meaning of a word is not.
Sources
- Knobe, J. (2003). Intentional action and side effects in ordinary language. Analysis, 63(3), 190–194. doi:10.1111/1467-8284.00419. Full text: course copy.
- Cova, F., Strickland, B., Abatista, A., et al. (2021). Estimating the reproducibility of experimental philosophy. Review of Philosophy and Psychology, 12, 9–44. doi:10.1007/s13164-018-0400-9. Accepted manuscript: UEA repository.
- Uttich, K., & Lombrozo, T. (2010). Norms inform mental state ascriptions: A rational explanation for the side-effect effect. Cognition, 116(1), 87–100. doi:10.1016/j.cognition.2010.04.003. Summary: Princeton Concepts and Cognition Lab.