Overconfidence

Overconfidence is the tendency to believe one knows more, or is more likely to be right, than the evidence warrants. The book treats it as a double-edged force: it makes experts and decision-makers misjudge their own reliability, but can also keep individuals from surrendering their private judgment to a crowd. Its consequences depend on whether confidence challenges an untested consensus or protects one from unwelcome evidence.

Experts and the Illusion of Calibration

Studies reviewed in the book show that experts’ judgments can be inconsistent both across people and within the same person. Agreement among specialists in fields including stock picking, livestock judging, and clinical psychology falls below 50 percent: experts in these areas are as likely to disagree as to agree. Medical pathologists, in one study, had an internal consistency of only 0.5, meaning that the same pathologist, presented with the same evidence, would give a different opinion half the time.

The problem is not only that judgments are unreliable; experts are also poor at gauging their own reliability. A survey found physicians, nurses, lawyers, engineers, entrepreneurs, and investment bankers believed they knew more than they did. Foreign-exchange traders overestimated the accuracy of their exchange-rate predictions 70 percent of the time. The book contrasts these examples with expert bridge players and weather forecasters, whose probability judgments are routinely well calibrated: when forecasters predict a 30 percent chance of rain, it rains on 30 percent of those days.

Expertise does not reliably announce itself through unusual confidence. Experts are no more confident in their abilities than average people, and self-assessment correlates very little with performance. Knowing something and knowing that one knows it are different skills. This weakens the appeal of finding a single expert whose certainty can stand in for evidence: confidence is no dependable badge of superior judgment.

Confidence, Crowds, and Cascades

The book’s account of collective judgment gives overconfidence a counterintuitive social role. A wise crowd depends on diversity of opinion, independent judgment, decentralization, and a way to aggregate individual estimates. When people make estimates from partly distinct information, their answers contain both information and error; averaging can cancel some errors while preserving the information. But aggregation is useful only when there is information in the judgments to begin with.

An information cascade develops when people give public information more weight than their own private evidence and follow what others have already done. Overconfident people are more likely to resist this pressure: they rely on their own instincts rather than treating the crowd’s behavior as decisive. Even when their confidence is misplaced and leads them personally to choose badly, their refusal to follow can disrupt the apparent signal that everyone else is receiving. That disruption can encourage others to reconsider their own information rather than imitate the group automatically.

This is not a general endorsement of overconfidence. Its social value is conditional: it can interrupt a harmful cascade when people have private information and when the available options and evidence are initially diverse. The book distinguishes this kind of dissent from imitation that follows useful information. Iowa farmers, for example, often waited to see neighbors succeed with hybrid corn before trying it, but they did not simply convert their entire farms at once. They tested the seed on a small part of a field and adopted it more broadly only after satisfying themselves. Half the farmers in the region took nine years to adopt it after the first farmer planted it. Their gradual learning was not blind conformity; it combined social evidence with independent testing.

The Columbia: Assurance Without Knowledge

The handling of the foam strike on the Columbia illustrates how overconfidence can turn uncertainty into organizational assurance. During the shuttle’s ascent, foam broke off the external fuel tank and struck the left wing. The Debris Assessment Team estimated the foam’s size and impact speed, then used the Crater algorithm to predict how far debris might penetrate the shuttle’s thermal-protection tiles. The team had no clear camera image of the impact, and its analysis remained inconclusive. Its members nevertheless told Don McCormack there was reason for concern.

In the Mission Management Team teleconference on January 21, McCormack did not convey the force of that concern. The foam strike appeared late in the meeting, after discussion of a jammed camera, scientific experiments, and a leaky water separator. Linda Ham, the team leader, replied that there was little that could be done and therefore the issue was not a factor during the flight. The book presents this response as part of a broader pattern: rather than beginning with evidence and working toward a conclusion, the team behaved as though the question of whether foam could cause serious damage had already been settled.

The refusal to authorize on-orbit pictures made the gap between assurance and knowledge consequential. Managers argued that the images would not have enough resolution to show the damaged area, although none had the security clearance needed to know how good the images would be, and none asked the Department of Defense, which would have taken them. The decision was thus made with little or no knowledge of the relevant imaging capability, but with an air of confidence. The book links this pattern to confirmation bias: once managers assumed that nothing serious was wrong, they discounted contrary evidence and treated further inquiry as unnecessary.

Caution Lost in the Decision Process

A January 24 briefing made the difference between the engineers’ uncertainty and management’s certainty especially plain. The Debris Assessment Team presented five possible scenarios and concluded that the shuttle was likely safe, but stressed the severe limits of its analysis. The engineers did not know exactly where the foam had struck because pictures had not been approved. Their Crater algorithm had been designed for debris hundreds of times smaller than the piece that hit the Columbia, so its result could not be trusted as a precise assessment of this impact.

Management focused on the reassuring portion of the conclusion. McCormack reported that the thermal analysis did not indicate a potential burn-through, while also acknowledging uncertainty about the debris’s size, impact location, and angle. Ham recast the exchange as confirmation that there was no catastrophic damage or safety-of-flight issue. The qualifications did not disappear from the record; they lost force in the group’s interpretation of it. What had been a tentative conclusion constrained by inadequate tools became, in the meeting, an apparent basis for taking no different action.

The episode exposes a distinction central to the book’s argument: confidence can be useful when it prevents people from automatically accepting the crowd’s signal, but dangerous when it closes a group’s inquiry before evidence has been gathered. In the Columbia case, uncertainty was present in the technical work and was voiced in the meetings; it was assurance, not a lack of warning, that governed how the information was handled. More broadly, overconfidence matters not simply because individuals make mistakes, but because a confident decision can determine which evidence an organization seeks, which doubts it preserves, and whether dispersed knowledge is allowed to affect a collective judgment.