Thinking, Fast and Slow (22 page)

A Defense of Extreme Predictions?

I introduced Tom W earlier to illustrate predictions of discrete outcomes such as field of specialization or success in an examination, which are expressed by assigning a probability to a specified event (or in that case by ranking outcomes from the most to the least probable). I also described a procedure that counters the common biases of discrete prediction: neglect of base rates and insensitivity to the quality of information.

The biases we find in predictions that are expressed on a scale, such as GPA or the revenue of a firm, are similar to the biases observed in judging the probabilities of outcomes.

The corrective procedures are also similar:

• Both contain a baseline prediction, which you would make if you knew nothing about the case at hand. In the categorical case, it was the base rate. In the numerical case, it is the average outcome in the relevant category.
  
• Both contain an intuitive prediction, which expresses the number that comes to your mind, whether it is a probability or a GPA.
  
• In both cases, you aim for a prediction that is intermediate between the baseline and your intuitive response.
  
• In the default case of no useful evidence, you stay with the baseline.
  
• At the other extreme, you also stay with your initial predictiononsр. This will happen, of course, only if you remain completely confident in your initial prediction after a critical review of the evidence that supports it.
  
• In most cases you will find some reason to doubt that the correlation between your intuitive judgment and the truth is perfect, and you will end up somewhere between the two poles.
  

This procedure is an approximation of the likely results of an appropriate statistical analysis. If successful, it will move you toward unbiased predictions, reasonable assessments of probability, and moderate predictions of numerical outcomes. The two procedures are intended to address the same bias: intuitive predictions tend to be overconfident and overly extreme.

Correcting your intuitive predictions is a task for System 2. Significant effort is required to find the relevant reference category, estimate the baseline prediction, and evaluate the quality of the evidence. The effort is justified only when the stakes are high and when you are particularly keen not to make mistakes. Furthermore, you should know that correcting your intuitions may complicate your life. A characteristic of unbiased predictions is that they permit the prediction of rare or extreme events only when the information is very good. If you expect your predictions to be of modest validity, you will never guess an outcome that is either rare or far from the mean. If your predictions are unbiased, you will never have the satisfying experience of correctly calling an extreme case. You will never be able to say, “I thought so!” when your best student in law school becomes a Supreme Court justice, or when a start-up that you thought very promising eventually becomes a major commercial success. Given the limitations of the evidence, you will never predict that an outstanding high school student will be a straight-A student at Princeton. For the same reason, a venture capitalist will never be told that the probability of success for a start-up in its early stages is “very high.”

The objections to the principle of moderating intuitive predictions must be taken seriously, because absence of bias is not always what matters most. A preference for unbiased predictions is justified if all errors of prediction are treated alike, regardless of their direction. But there are situations in which one type of error is much worse than another. When a venture capitalist looks for “the next big thing,” the risk of missing the next Google or Facebook is far more important than the risk of making a modest investment in a start-up that ultimately fails. The goal of venture capitalists is to call the extreme cases correctly, even at the cost of overestimating the prospects of many other ventures. For a conservative banker making large loans, the risk of a single borrower going bankrupt may outweigh the risk of turning down several would-be clients who would fulfill their obligations. In such cases, the use of extreme language (“very good prospect,” “serious risk of default”) may have some justification for the comfort it provides, even if the information on which these judgments are based is of only modest validity.

For a rational person, predictions that are unbiased and moderate should not present a problem. After all, the rational venture capitalist knows that even the most promising start-ups have only a moderate chance of success. She views her job as picking the most promising bets from the bets that are available and does not feel the need to delude herself about the prospects of a start-up in which she plans to invest. Similarly, rational individuals predicting the revenue of a firm will not be bound to a singleys р number—they should consider the range of uncertainty around the most likely outcome. A rational person will invest a large sum in an enterprise that is most likely to fail if the rewards of success are large enough, without deluding herself about the chances of success. However, we are not all rational, and some of us may need the security of distorted estimates to avoid paralysis. If you choose to delude yourself by accepting extreme predictions, however, you will do well to remain aware of your self-indulgence.

Perhaps the most valuable contribution of the corrective procedures I propose is that they will require you to think about how much you know. I will use an example that is familiar in the academic world, but the analogies to other spheres of life are immediate. A department is about to hire a young professor and wants to choose the one whose prospects for scientific productivity are the best. The search committee has narrowed down the choice to two candidates:

The intuitive choice favors Kim, because she left a stronger impression, and WYSIATI. But it is also the case that there is much less information about Kim than about Jane. We are back to the law of small numbers. In effect, you have a smaller sample of information from Kim than from Jane, and extreme outcomes are much more likely to be observed in small samples. There is more luck in the outcomes of small samples, and you should therefore regress your prediction more deeply toward the mean in your prediction of Kim’s future performance. When you allow for the fact that Kim is likely to regress more than Jane, you might end up selecting Jane although you were less impressed by her. In the context of academic choices, I would vote for Jane, but it would be a struggle to overcome my intuitive impression that Kim is more promising. Following our intuitions is more natural, and somehow more pleasant, than acting against them.

You can readily imagine similar problems in different contexts, such as a venture capitalist choosing between investments in two start-ups that operate in different markets. One start-up has a product for which demand can be estimated with fair precision. The other candidate is more exciting and intuitively promising, but its prospects are less certain. Whether the best guess about the prospects of the second start-up is still superior when the uncertainty is factored in is a question that deserves careful consideration.

A Two-Systems View of Regression

Extreme predictions and a willingness to predict rare events from weak evidence are both manifestations of System 1. It is natural for the associative machinery to match the extremeness of predictions to the perceived extremeness of evidence on which it is based—this is how substitution works. And it is natural for System 1 to generate overconfident judgments, because confidence, as we have seen, is determined by the coherence of the best story you can tell from the evidence at hand. Be warned: your intuitions will deliver predictions that are too extreme and you will be inclinehe рd to put far too much faith in them.

Regression is also a problem for System 2. The very idea of regression to the mean is alien and difficult to communicate and comprehend. Galton had a hard time before he understood it. Many statistics teachers dread the class in which the topic comes up, and their students often end up with only a vague understanding of this crucial concept. This is a case where System 2 requires special training. Matching predictions to the evidence is not only something we do intuitively; it also seems a reasonable thing to do. We will not learn to understand regression from experience. Even when a regression is identified, as we saw in the story of the flight instructors, it will be given a causal interpretation that is almost always wrong.

Speaking of Intuitive Predictions

The trader-philosopher-statistician Nassim Taleb could also be considered a psychologist. In
The Black Swan
, Taleb introduced the notion of a
narrative fallacy
to describe how flawed stories of the past shape our views of the world and our expectations for the future. Narrative fallacies arise inevitably from our continuous attempt to make sense of the world. The explanatory stories that people find compelling are simple; are concrete rather than abstract; assign a larger role to talent, stupidity, and intentions than to luck; and focus on a few striking events that happened rather than on the countless events that failed to happen. Any recent salient event is a candidate to become the kernel of a causal narrative. Taleb suggests that we humans constantly fool ourselves by constructing flimsy accounts of the past and believing they are true.

Good stories provide a simple and coherent account >

A compelling narrative fosters an illusion of inevitability. Consider the story of how Google turned into a giant of the technology industry. Two creative graduate students in the computer science department at Stanford University come up with a superior way of searching information on the Internet. They seek and obtain funding to start a company and make a series of decisions that work out well. Within a few years, the company they started is one of the most valuable stocks in America, and the two former graduate students are among the richest people on the planet. On one memorable occasion, they were lucky, which makes the story even more compelling: a year after founding Google, they were willing to sell their company for less than $1 million, but the buyer said the price was too high. Mentioning the single lucky incident actually makes it easier to underestimate the multitude of ways in which luck affected the outcome.

A detailed history would specify the decisions of Google’s founders, but for our purposes it suffices to say that almost every choice they made had a good outcome. A more complete narrative would describe the actions of the firms that Google defeated. The hapless competitors would appear to be blind, slow, and altogether inadequate in dealing with the threat that eventually overwhelmed them.

I intentionally told this tale blandly, but you get the idea: there is a very good story here. Fleshed out in more detail, the story could give you the sense that you understand what made Google succeed; it would also make you feel that you have learned a valuable general lesson about what makes businesses succeed. Unfortunately, there is good reason to believe that your sense of understanding and learning from the Google story is largely illusory. The ultimate test of an explanation is whether it would have made the event predictable in advance. No story of Google’s unlikely success will meet that test, because no story can include the myriad of events that would have caused a different outcome. The human mind does not deal well with nonevents. The fact that many of the important events that did occur involve choices further tempts you to exaggerate the role of skill and underestimate the part that luck played in the outcome. Because every critical decision turned out well, the record suggests almost flawless prescience—but bad luck could have disrupted any one of the successful steps. The halo effect adds the final touches, lending an aura of invincibility to the heroes of the story.

Like watching a skilled rafter avoiding one potential calamity after another as he goes down the rapids, the unfolding of the Google story is thrilling because of the constant risk of disaster. However, there is foр an instructive difference between the two cases. The skilled rafter has gone down rapids hundreds of times. He has learned to read the roiling water in front of him and to anticipate obstacles. He has learned to make the tiny adjustments of posture that keep him upright. There are fewer opportunities for young men to learn how to create a giant company, and fewer chances to avoid hidden rocks—such as a brilliant innovation by a competing firm. Of course there was a great deal of skill in the Google story, but luck played a more important role in the actual event than it does in the telling of it. And the more luck was involved, the less there is to be learned.

At work here is that powerful WY SIATI rule. You cannot help dealing with the limited information you have as if it were all there is to know. You build the best possible story from the information available to you, and if it is a good story, you believe it. Paradoxically, it is easier to construct a coherent story when you know little, when there are fewer pieces to fit into the puzzle. Our comforting conviction that the world makes sense rests on a secure foundation: our almost unlimited ability to ignore our ignorance.

I have heard of too many people who “knew well before it happened that the 2008 financial crisis was inevitable.” This sentence contains a highly objectionable word, which should be removed from our vocabulary in discussions of major events. The word is, of course,
knew
. Some people thought well in advance that there would be a crisis, but they did not know it. They now say they knew it because the crisis did in fact happen. This is a misuse of an important concept. In everyday language, we apply the word
know
only when what was known is true and can be shown to be true. We can know something only if it is both true and knowable. But the people who thought there would be a crisis (and there are fewer of them than now remember thinking it) could not conclusively show it at the time. Many intelligent and well-informed people were keenly interested in the future of the economy and did not believe a catastrophe was imminent; I infer from this fact that the crisis was not knowable. What is perverse about the use of
know
in this context is not that some individuals get credit for prescience that they do not deserve. It is that the language implies that the world is more knowable than it is. It helps perpetuate a pernicious illusion.

The core of the illusion is that we believe we understand the past, which implies that the future also should be knowable, but in fact we understand the past less than we believe we do.
Know
is not the only word that fosters this illusion. In common usage, the words
intuition
and
premonition
also are reserved for past thoughts that turned out to be true. The statement “I had a premonition that the marriage would not last, but I was wrong” sounds odd, as does any sentence about an intuition that turned out to be false. To think clearly about the future, we need to clean up the language that we use in labeling the beliefs we had in the past.

The Social Costs of Hindsight

The mind that makes up narratives about the past is a sense-making organ. When an unpredicted event occurs, we immediately adjust our view of the world to accommodate the surprise. Imagine yourself before a football game between two teams that have the same record of wins and losses. Now the game is over, and one team trashed the other. In your revised model of the world, the winning team is much stronger than the loser, and your view of the past as well as of the future has been altered be fрy that new perception. Learning from surprises is a reasonable thing to do, but it can have some dangerous consequences.

A general limitation of the human mind is its imperfect ability to reconstruct past states of knowledge, or beliefs that have changed. Once you adopt a new view of the world (or of any part of it), you immediately lose much of your ability to recall what you used to believe before your mind changed.

Many psychologists have studied what happens when people change their minds. Choosing a topic on which minds are not completely made up—say, the death penalty—the experimenter carefully measures people’s attitudes. Next, the participants see or hear a persuasive pro or con message. Then the experimenter measures people’s attitudes again; they usually are closer to the persuasive message they were exposed to. Finally, the participants report the opinion they held beforehand. This task turns out to be surprisingly difficult. Asked to reconstruct their former beliefs, people retrieve their current ones instead—an instance of substitution—and many cannot believe that they ever felt differently.

Your inability to reconstruct past beliefs will inevitably cause you to underestimate the extent to which you were surprised by past events. Baruch Fischh off first demonstrated this “I-knew-it-all-along” effect, or
hindsight bias
, when he was a student in Jerusalem. Together with Ruth Beyth (another of our students), Fischh off conducted a survey before President Richard Nixon visited China and Russia in 1972. The respondents assigned probabilities to fifteen possible outcomes of Nixon’s diplomatic initiatives. Would Mao Zedong agree to meet with Nixon? Might the United States grant diplomatic recognition to China? After decades of enmity, could the United States and the Soviet Union agree on anything significant?

After Nixon’s return from his travels, Fischh off and Beyth asked the same people to recall the probability that they had originally assigned to each of the fifteen possible outcomes. The results were clear. If an event had actually occurred, people exaggerated the probability that they had assigned to it earlier. If the possible event had not come to pass, the participants erroneously recalled that they had always considered it unlikely. Further experiments showed that people were driven to overstate the accuracy not only of their original predictions but also of those made by others. Similar results have been found for other events that gripped public attention, such as the O. J. Simpson murder trial and the impeachment of President Bill Clinton. The tendency to revise the history of one’s beliefs in light of what actually happened produces a robust cognitive illusion.

Hindsight bias has pernicious effects on the evaluations of decision makers. It leads observers to assess the quality of a decision not by whether the process was sound but by whether its outcome was good or bad. Consider a low-risk surgical intervention in which an unpredictable accident occurred that caused the patient’s death. The jury will be prone to believe, after the fact, that the operation was actually risky and that the doctor who ordered it should have known better. This outcome bias makes it almost impossible to evaluate a decision properly—in terms of the beliefs that were reasonable when the decision was made.

Hindsight is especially unkind to decision makers who act as agents for others—physicians, financial advisers, third-base coaches, CEOs, social workers, diplomats, politicians. We are prone to blame decision makers for good decisions that worked out badly and to give them too little credit for successful movesecaр that appear obvious only after the fact. There is a clear
outcome bias
. When the outcomes are bad, the clients often blame their agents for not seeing the handwriting on the wall—forgetting that it was written in invisible ink that became legible only afterward. Actions that seemed prudent in foresight can look irresponsibly negligent in hindsight. Based on an actual legal case, students in California were asked whether the city of Duluth, Minnesota, should have shouldered the considerable cost of hiring a full-time bridge monitor to protect against the risk that debris might get caught and block the free flow of water. One group was shown only the evidence available at the time of the city’s decision; 24% of these people felt that Duluth should take on the expense of hiring a flood monitor. The second group was informed that debris had blocked the river, causing major flood damage; 56% of these people said the city should have hired the monitor, although they had been explicitly instructed not to let hindsight distort their judgment.

The worse the consequence, the greater the hindsight bias. In the case of a catastrophe, such as 9/11, we are especially ready to believe that the officials who failed to anticipate it were negligent or blind. On July 10, 2001, the Central Intelligence Agency obtained information that al-Qaeda might be planning a major attack against the United States. George Tenet, director of the CIA, brought the information not to President George W. Bush but to National Security Adviser Condoleezza Rice. When the facts later emerged, Ben Bradlee, the legendary executive editor of
The Washington Post
, declared, “It seems to me elementary that if you’ve got the story that’s going to dominate history you might as well go right to the president.” But on July 10, no one knew—or could have known—that this tidbit of intelligence would turn out to dominate history.

Because adherence to standard operating procedures is difficult to second-guess, decision makers who expect to have their decisions scrutinized with hindsight are driven to bureaucratic solutions—and to an extreme reluctance to take risks. As malpractice litigation became more common, physicians changed their procedures in multiple ways: ordered more tests, referred more cases to specialists, applied conventional treatments even when they were unlikely to help. These actions protected the physicians more than they benefited the patients, creating the potential for conflicts of interest. Increased accountability is a mixed blessing.

Although hindsight and the outcome bias generally foster risk aversion, they also bring undeserved rewards to irresponsible risk seekers, such as a general or an entrepreneur who took a crazy gamble and won. Leaders who have been lucky are never punished for having taken too much risk. Instead, they are believed to have had the flair and foresight to anticipate success, and the sensible people who doubted them are seen in hindsight as mediocre, timid, and weak. A few lucky gambles can crown a reckless leader with a halo of prescience and boldness.

Recipes for Success

The sense-making machinery of System 1 makes us see the world as more tidy, simple, predictable, and coherent than it really is. The illusion that one has understood the past feeds the further illusion that one can predict and control the future. These illusions are comforting. They reduce the anxiety that we would experience if we allowed ourselves to fully acknowledge the uncertainties of existence. We all have a need for the reassuring message that actions have appropriate consequences, and that success will reward wisdom and courage. Many bdecрusiness books are tailor-made to satisfy this need.