Kahneman asks us to consider a scenario and note our intuitive answer: a cab was involved in a hit and run accident at night. 85% of the cabs in the city are green and 15% are blue. A witness identified the cab as blue; witnesses under these circumstances correctly identify cab colors 80% of the time. Usually, people ignore the base rates of the number of cabs, and instead favor the witness’s accuracy, guessing about 80% percent. According to Bayes’s rule, the correct answer is 41%.
This chapter becomes another prime example to support the idea that people vastly prefer stories over statistics. Instead of relying on the numbers provided to them (the fact that there are far more green cabs than blue cabs), people prefer to rely on the story provided by the witness.
However, if the first sentence had said that green cabs are involved in 85% of accidents, people give more weight to that information because they construct a story assuming that the green cabs are more reckless. The information is literally the same, but people prefer base rates that hint at a cause.
This slight change in the way the information is presented also supports the idea that stories take precedence over statistics, as this fact provides people with the ability to construct a coherent story about the green cabs.
The causal version of the cab problem creates a stereotype that green cab drivers are dangerous. Kahneman admits that social stereotypes can be harmful, but that stereotypes in general allow us to create categories and norms—like horses, refrigerators, and police officers. Stereotyping in the case of the green cabs makes people more accurate.
A stereotype is, at its core, a way of constructing stories about a given category in order to make the world more coherent and reliant on patterns—which is why our System 1 likes to rely on them.
Kahneman and Tversky borrowed the notion of causal base rates from Icek Ajzen. In an experiment, Ajzen showed participants descriptions of students. He told one group that 75% of the students passed an exam, and told another group that 25% of the students had passed. Every student was judged more likely to pass the high-success condition than in the high-failure rate, because participants assumed that the test had been brutally difficult.
Like the green cab experiment, constructing a stereotype allowed the participants to make correct inferences about the students—that generally, it is safe to assume that the students in the 75% passing group were more likely to have done better than the students in the 25% passing group. But the next example complicates the idea that people understand this concept.
In another classic experiment, social psychologists Richard Nisbett and Eugene Borgida told their students about a “helping experiment” that had been conducted a few years earlier. Participants were separated into individual booths made to think that someone in another booth was having a seizure and choking. Only four out of fifteen them responded to the person’s call for help.
Nisbett and Borgida make very clear the base-rates of their experiment. Even though we might like to believe otherwise, eleven out of fifteen people will not rush to help a dying stranger if they believe that someone else has heard the same call.
Nisbett and Borgida described this experiment in the hopes that their students would see the low base-rates and assume that it was a difficult test. But when students were shown videos of brief interviews with two of the participants, who appear to be nice, normal decent people, the students believed that both individuals would rush to the choking person’s aid—despite the fact that they knew there was only a 27% chance of this being the case.
The 27% statistic is surprising, and it conflicts with our idea of people (and of ourselves) as generally decent and helpful. And so, when individuals appear to be decent and helpful, they confirm our previously held beliefs and this information takes precedence over the statistic.
For a teacher of psychology, Kahneman writes, the study is disheartening because the results did not change their beliefs about people’s behavior. But in another part of the experiment, Nisbett showed another group of students the two interviews (without describing the full results) and told his students that these two individuals did not help the choking person. Nesbitt and Borgia then asked them to guess the global results, and the students’ guesses were extremely accurate.
Again, the stories (particularly about individuals) take precedence over the statistics. When confronted with surprising individual cases, we are more likely to make accurate inferences about the general population than if they are shown a surprising statistic and unsurprising individual cases.
The results demonstrate that when students were surprised by a statistical fact, the students did not change their assumptions. But when surprised by individual cases, they immediately made the generalization and inferred that helping is more difficult than they thought. This, Kahneman says, is why his book contains questions that are addressed to the reader: being surprised by one’s own behavior is more powerful than being surprised by people’s behavior more generally.
Kahneman’s style in the book—which often uses “you” and “we” pronouns, takes its basis from this principle. We often believe that certain psychological principals don’t apply to us, and when we are surprised by individual cases (including ourselves), we are more likely to learn the general lessons that he offers.