BRAIN WORLD: “Intuition” is a vague term. How would you define it?
DANIEL KAHNEMAN: I haven’t always been consistent in my use of the term. The first time that I used it in print, I think, was in a paper with Amos Tversky on the “Law of Small Numbers.” In that paper, we wrote something to the effect that “the Law of Large Numbers is not part of the repertoire of human intuition.” What we meant there by “intuition” was a descriptive generalization made by an objective observer about the rule that appears to generate the judgments of an individual about specific cases. And this can be an accurate description by the outside observer, even if the individual would reject this rule if he were to consider it.
I have also used the term to describe an intuitive system, a way of generating thoughts that are called intuitive thoughts. The operations of this system are typically automatic, quick, effortless, associative and often emotionally charged. Usually, they are not open to introspection, and difficult to control or change. This intuitive system generates involuntary impressions that come to mind spontaneously, like percepts. Intuitive judgments directly reflect these impressions with little modification from the reasoning system, which functions very differently from the intuitive system.
BW: There appears to be a difference between what you call intuitive judgments, understood as a product of the intuitive system, and the type of case judgments that philosophers use to build and test their moral theories, which they also sometimes refer to as “intuitions.” Philosophers typically take these to be considered case judgments, that is, judgments that arise after reflection on the case and on our reasons for judging it as we do, whereas your category of intuitive judgments does not appear to involve such extensive reflection. What do you think of this method of arriving at considered case judgments?
DK: Well, there is a very interesting contrast between the way a psychologist would go at it and the way a philosopher would go at it, and it has a lot to do with the idea that the philosopher can know why she is reaching this conclusion. A basic assumption in psychological analysis is that you first have an intuition in a given situation. Then, when you ask yourself why you have it, you tell yourself a story. But the story you come up with does not necessarily identify the cause of your intuitions, because you typically do not have access to what causes your intuitions.
You know, many psychologists believe that consciousness is a story that we tell ourselves about ourselves. And in many cases, the story does not correspond to reality. I mean, it is very easy to create cases where you know the story isn’t true. Let me give you an example, just to push the point a little. There are experiments with posthypnotic suggestion, where you tell somebody, “I’ll clap my hands and then you’ll get up and open the window.” The person wakes up, you clap your hands, and he gets up and opens the window. If you ask him, “Why did you open the window?” he’ll say something like, “The room felt very warm.”
BW: But aren’t these cases of hypnosis simply pathological?
DK: No, they are the best examples! The beauty of these examples is that you know why the subjects are doing it. They are doing it because they were given the instruction, and then somebody clapped his hands. But they have a completely different experience of why they are doing it. What’s more, people are never at a loss for a reason why they did it. They can be made to do absolutely absurd things with posthypnotic suggestion, and yet these things seem to make sense to them when they do them. The conclusion I draw from this is that the mental operation of making sense of our intuitive judgments is a very different cognitive activity from having these intuitions.
BW: Still, the case of the hypnotic seems to show only that we cannot always trust our introspective judgments about why we act or judge as we do; it doesn’t show that we can never trust these judgments. And the method employed … others involve more than simple introspective judgment about what is driving our judgments in isolated cases. Philosophers employing this method go on to consider whether the reason they think they might be responding to in a particular case determines their judgment in other cases also.
DK: But this raises another methodological problem. There is a distinction between what I have called the “within-subject” and the “between-subject” method of considering cases. The within-subject method involves noting the intuitions a given subject has when she is considering multiple contrasting cases. The between-subject method involves noting the intuitions a subject has in a single case without engaging in explicit comparison with other cases. The two methods can elicit very different intuitions about the cases considered.
Now, because its aim is to find rules that are applicable across different cases, moral philosophy is, by its nature, restricted to the within-subject method. That is, a moral philosopher is always conscious of two or more cases and has intuitions about the differences in the two cases and about whether these differences are relevant to how she should respond. And I think this is a major limitation of moral philosophy, because, in their everyday life, people are confronted with problems one at a time, so their relevant intuitions are about cases that occur one at a time. And the moral philosopher’s stance prevents her from identifying the moral intuitions that are relevant to individuals who live their lives in this way. For this purpose, the between-subject approach,which involves asking one group of people a bout one case and another group of people about another case, is superior.
BW: That’s very interesting … I suppose I see how people’s tendency to respond differently to the same case depending on whether it is framed as involving gains or losses … would only clearly emerge in a between-subject experiment. Can you give some other examples of cases in which the two methods lead to different judgments?
DK: Let me give you two examples. First, the intuitive judgments in the grocery store cases I mentioned before, which revealed the effect of poignancy on financial compensation for the victim, could only have been elicited by the between-subject method. For if one considers the two cases side-by-side, whether the victim was at his regular grocery store or not would have appeared irrelevant, and so would have made no difference to the compensation awarded.
Second, in research I carried out with Cass Sunstein, David Schkade and Ilana Ritov, we found that the punishments people thought appropriate when considering in isolation cases of business fraud differed significantly from punishments they thought appropriate when they compared cases of business fraud with wrongdoing involving bodily injury. Our hypothesis was that outrage at an act was a significant contributor to the size of the punishment people judged appropriate. When they considered a case of business fraud in isolation, people implicitly compared the case to other cases of business fraud, so their level of outrage was determined by how egregious the conduct was as compared to other cases of business fraud. A particularly egregious case of business fraud would therefore be very heavily punished.
However, in comparing cases of business fraud with cases of a different category altogether, like bodily injury, the relative importance of the category to which the wrongdoing belongs became relevant as well. Since bodily injury was regarded as generally worse than fraud, this could make the same case of fraud seem less worthy of severe punishment.
BW: But these cases seem to vindicate rather than undermine the within-subject approach used by philosophers. For in these cases, the within-subject approach of having a single person consider multiple cases side by side seems to prevent mistakes that would occur if we considered each case separately.
DK: Well, to some extent I agree, and in that paper with Sunstein, Schkade and Ritov, we indeed try to argue that the within-subject comparison has an advantage because it gives you a better shot at approaching a consistent view of the world than judging single cases in isolation would, even though it doesn’t always work. But when we are thinking about policies and about applying moral rules to people, then it’s important to remember that these people will experience cases as participants in the between-subject experiment.
What is morally compelling to us as we consider the contrast between the two cases may not seem morally relevant to them at all, and so the rules we may wish to impose on them may not seem relevant to them.
There is, I think, a genuine dilemma here between the demands of coherence and the need not to impose on people principles that violate their judgments. A solution that we arrived at in that paper is to partially accommodate both demands. So our recommendation was to measure outrage at the wrongful behavior, considered in isolation, and use outrage as an input in determining the severity of punishment. We would not recommend allowing outrage alone to determine punishment, as it appears to do in some jury trials.
Our thinking was that while a public policy that merely reflects outrage is grotesque, a policy that is insensitive to outrage is not going to be acceptable to people. So public policy should be sensitive to outrage, but not dominated by it. That is the best we could do. It is not a unique solution, but it acknowledges that moral intuition about a specific case can neither be trusted nor altogether ignored.