Volume 10, No. 3 - April 2000
More in this Issue

Can you spot a liar? Most people think they can, but they're wrong. In study after study, subjects have been unable to distinguish between videotaped liars and truth tellers, scoring little better than chance. That goes even for those who feel especially sure they can catch liars--police detectives, for instance.

Human beings, as it turns out, have too much faith in themselves. The detection of falsehood is scarcely the only domain where they overestimate their abilities. A survey of British motorists not long ago revealed that 95 percent thought they were better-than-average drivers. Similarly, most people think they are likely to live longer than the mean. In a classic 1977 paper in the Journal of Experimental Psychology, Baruch Fischhoff, Paul Slovic, and Sarah Lichtenstein reported that people often pronounce themselves absolutely certain of beliefs that are untrue. Subjects would declare themselves 100 percent sure that, say, the potato originated in Ireland, when it actually came from Peru.

Overconfidence is nearly universal. In fact, a study some years ago found that the only group of people free from it--the only group with a realistic view of their own capacities--were the clinically depressed. But is overconfidence distributed equally? Not according to a widely publicized paper in last December's issue of the Journal of Personality and Social Psychology. The authors, David A. Dunning of Cornell and his graduate student Justin Kruger, drew a poignant conclusion from their research: The most incompetent people have the most inflated notion of their abilities. "Not only do they reach erroneous conclusions and make unfortunate choices," the two psychologists wrote, "but their incompetence robs them of the ability to realize it."

Dunning and Kruger administered three tests, measuring ability in logic, English grammar, and humor appreciation (the subjects' ratings of jokes were judged against those of a panel of professional comedians). In all three areas, the subjects who did worst were the most likely to "grossly overestimate" how well they had performed. Those who scored in the twelfth percentile in the logic test, for example, imagined that their skill in logic was at the sixty-eighth percentile.

Now, if you are among the competent, you might derive some consolation from this research, since it implies that you are unlikely to be grossly overconfident. But perhaps you simply imagine that you are among the competent--precisely because you suffer from the overconfidence of the incompetent. And there is more to worry about. Overconfidence may decrease with competence, but other studies show that it increases with knowledgeability; that is, the more specialized information you have about something, the more likely you are to be overconfident in your judgments about it. Overconfidence also tends to rise with the complexity of the problem. This means that experts reasoning about difficult matters--doctors, engineers, financial analysts, academics, even the pope when he is not speaking ex cathedra--are apt to be seriously overconfident in the validity of their conclusions.

Let me illustrate the point with an anecdote. Paul Erdos was one of the supreme mathematicians of the last century. He was also one of the world's leading experts on probability theory. Indeed, he invented the probabilistic method, which is often simply referred to as the Erdos method--thus making his name synonymous with probability (and an apparent exception to Stigler's law [see last month's Hypotheses]).

In 1991, Erdos found himself befuddled when the Parade magazine columnist Marilyn vos Savant published a probability puzzle called the Monty Hall problem, named after the emcee of the defunct game show Let's Make a Deal. It goes like this: There are three doors on stage, labeled A, B, and C. Behind one of them is a sports car; behind the other two are goats. You get to choose one of the doors and keep whatever is behind it. Let's suppose you choose door A. Now, instead of showing you what's behind door A, Monty Hall slyly opens door B and reveals ... a goat. He then offers you the option of switching to door C. Should you take it? (Assume, for the sake of argument, that you are indifferent to the charm of goats.)

Counterintuitively enough, the answer is that you should switch, since a switch increases your chance of winning from one-third to two-thirds. Why? When you initially chose door A, there was a one-third chance you would win the car. Monty's crafty revelation that there's a goat behind door B gives no new information about what's behind the door you already chose--you already know one of the other two doors has to conceal a goat--so the likelihood that the car is behind door A remains one-third. Which means that, with door B eliminated, there is a two-thirds chance that the car is behind door C.

But Erdos insisted to his friends it wasn't so. His intuition told him that switching should make no difference in the odds. And this peerless authority on probability was confident in his intuition--so confident that he remained in high dudgeon for several days, until a mathematician at Bell Labs finally made him see his error. (This episode is re counted in recent biographies by Bruce Schechter and Paul Hoffman. Any reader who is still reluctant to switch can find a fuller discussion at www.linguafranca.com.)

One final generalization can be drawn from the psychological literature: High levels of confidence are usually associated with high levels of overconfidence. The gap between conviction and truth seems to be greatest in the case of those judgments about which one feels most certain. Who knows? The most overconfident judgment in history might turn out to be cogito ergo sum.

In addition to "Hypotheses," Jim Holt writes a regular column for the Wall Street Journal covering books on science and philosophy. His most recent book is Worlds Within Worlds: How the Infitesimal Revolutionized Thought (Four Walls/Eight Windows).