Thinking, Fast and Slow

Chapter 20 explains a phenomenon that Kahneman discovered and named the illusion of validity. He recounts an experience from his time in the psychology department of the Israeli army, where he was occasionally charged with evaluating cadets for potential officer training. After several efforts, it was clear that the group had developed no useful means of predicting leadership ability. Their predictions were no better than random guesses, which was readily apparent. Nonetheless, they continued to act and feel that their recommendations were valid.

Kahneman notes the various factors at play, such as WYSIATI, and remarks that his experience was much like the students who believed that both subjects they evaluated were likely to be among the four of 15 people who immediately aided another subject when he appeared to have a seizure. Kahneman also provides some discussion of the financial world, in which stock-picking operates primarily under an illusion of validity. Pundits’ predictions are similarly treated as valid. In all these cases, Kahneman points to the existence of a professional culture that feeds the illusion of validity and observes that making accurate predictions is extremely difficult. Some short-term forecasting is possible, but true long-term prediction is not. The line between the two remains unknown.

Kahneman further develops and explains his view that, in most situations, decisions are better made by following a formula than by following expert opinions. He discusses psychologist Paul Meehl, whose work comparing clinical predictions against those generated by statistics shocked many clinical psychologists and remains significant more than 50 years later. Kahneman then points to specific contexts, such as the wine industry and medical school admissions decisions, to explain why formulas are more successful than expert intuition.

Kahneman asserts that reliance on formulas could improve predictive results in settings where decisions are currently made in low-validity environments. Interviews of medical school candidates probably reduce the ability to predict the best students, Kahneman suggests, because the faculty are inclined to rely on their opinion of the individual with whom they met. That type of reliance on personal judgment is vulnerable to the vagaries of System 1. Similarly, in the wine industry, tasting by experts likely undermines predictive value of their opinions that are otherwise based on factors like weather and region.

Perhaps the most potent example of the value of algorithms (or formulas) is the Apgar test regarding the health of newborn babies. As told by Kahneman, a resident asked Dr. Apgar how to make a systemic assessment of a newborn. She responded by writing down the five variables and three possible scores per variable, at which point she realized the value such a consistent and simple formula could have. The Apgar test is now standard in delivery rooms and is credited with reducing infant mortality by improving the accuracy of health assessments.

The heart of the issue regarding formulas versus expert opinion relates to the scope of the prediction needed. He suggests that Meehl’s work became controversial among clinical psychologists, especially those who likely overestimated their skill in making long-term predictions (such as the lifetime outcomes for a patient) based on the quality of their short-term assessments (such as a patient’s response to particular statements or approaches during therapy).

Kahneman recounts that he was assigned to design an interview for the Israeli army in 1955 (at age 22 and with only a bachelor’s degree) that would determine where incoming cadets were assigned. He replaced a system of intuitive judgment with one based on Meehl’s book (which had come out just a year earlier), such that interviewers were instructed to ask a series of preset questions, score the results according to directions, and then feed that raw data into a formula. The interviewers were upset with the change, so Kahneman instructed that, once the objective data was collected and recorded, they should close their eyes and assign a value to their intuitive assessment of the interviewee. Kahneman’s changes were a vast improvement, but he also learned something unexpected: the values assigned by the interviewers based on their intuitive judgment were as accurate as the overall values from the structured interviews.

Kahneman does not discount the value of intuitive judgments, but he strongly cautions that they should be used in conjunction with formulaic assessment techniques that generate objective data. He attributes the interviewer’s much-improved intuitive judgments to their role gathering the same objective data about individuals before voicing their intuitive conclusions as a value to be considered.

These lessons can benefit anyone who conducts interviews to make recruitment decisions, as the end of the chapter makes plain. Further, Kahneman explains that to avoid halo effects, it is essential to ask and score each question (or set of questions) in order before moving to the next item.

Chapter 22 discusses Kahneman’s productive animosity with Gary Klein, which eventually became a friendship. Klein strongly defended the importance of intuitive judgments. He came to study intuitive judgments by studying firefighting team leaders, who make decisions without comparing options (they generate one option and mentally simulate it to check for shortcomings). The process requires extensive experience to draw from memory and involves System 2 as well as System 1.

Intuition honed by expertise requires extensive skill in a field, which takes time and effort to develop. Kahneman and Klein had worked with different types of experts. Klein’s subjects had real expertise—fire commanders, clinical nurses, and the like—whereas Kahneman became familiar with stock-pickers, political economists, and clinicians attempting to make long-term predictions.

Klein and Kahneman came to agree that someone’s confidence in their own judgment is not a reliable guide to their accuracy. Expertise that can be trusted requires development of a skill, which itself requires two things: an environment of such sufficient regularity that scenarios can be predicted, and extensive experience that allows one to become well acquainted with recurring features of the environment. Regularity varies across environments. Chess, for example, is extremely regular.

The opposite extreme may be environments described as “wicked”—those where the wrong lessons are likely to be learned through practice. For example, a physician once believed he had become expert at determining who would soon develop typhoid, but he had simply failed to wash his hands before palpating the tongue of one patient after another. He correctly predicted that many people would develop typhoid, but it was his unknowing transmission of the condition, not professional intuition, that explained his apparent “success” in predicting typhoid development.

The clinicians challenged so strongly by Meehl were not incompetent; they just faced very difficult long-term tasks. Algorithms produced somewhat more accurate results but continue to have large error rates. Perhaps the fault lies in those supposed experts who claim to be able to predict the unpredictable, not in the flaws of intuition.

Kahneman settles on the nature of the environment and challenge as likely determinative. The availability of accurate feedback is essential for learning, as is the time necessary to practice and improve. The validity of expert intuition, Kahneman and Klein ultimately agreed, depends on whether the expert has had the opportunity to develop adequate experience operating in an environment of sufficient regularity. System 1 can and will detect irregularities, but it will create great confidence about false answers as well. A true expert’s intuition is likely worth trusting, but any such expert has previously made major mistakes in their field with great confidence.

Chapter 23 tells the reader that evaluations are best performed by those who can take “the outside view.” Otherwise, those evaluating their own group’s potential to meet its goals are likely to suffer from what Kahneman and Tversky would dub “the planning fallacy,” which occurs when there are forecasts that (1) trend unrealistically close to the best-case option and (2) would likely be enhanced by reviewing similar cases.

Essentially, this chapter details how actors who are heavily involved with a group or project can, hopefully, avoid losing perspective on their projects in a larger sense. The key is called “reference class forecasting.” This involves identifying a reference class, obtaining statistics of the class sufficient to generate a baseline prediction, and using specific information about the project to evaluate it against the class.

Kahneman discusses some of the paradoxical features of optimism, focusing on entrepreneurs as his prime example. Kahneman notes that most people have an optimism bias, then explains that optimism seems to propel optimists to believe they are special. They are, in fact, more likely to pull off their overly ambitious agendas than the alternative type. Without ever noting an alternative, Kahneman directly states: “When action is needed, optimism, even of the mildly delusional variety, may be a good thing” (257). This comment is followed by information that suggests “optimism is widespread, stubborn, and costly” (257).

Optimism is driven by overconfidence. Its main benefit is persistence despite setbacks. Whether this is a net benefit is unclear. What is clear is that most people (perhaps 90%) think themselves above average on desirable traits. Overly optimistic people take risks that most would avoid, perhaps because they do not believe the usual result is relevant. Yet, with regard to implementing ideas, such optimists carry forward.

Kahneman states that he is “not optimistic” about the ability to tame excessive optimism. He points to Klein’s work as suggesting organizations may be able do so more effectively than individuals. Klein suggests that organizations undertake a “premortem.” When a group has essentially made a major decision but not fully committed to it, knowledgeable individuals within the organization pretend that it is a year later and the decision proved disastrous, then write a brief history of the imagined disaster. Doing so reduces the impact of WYSIATI and overconfidence.