National Aeronautics and Space Administration

NASA’s account in The Wisdom of Crowds is defined by a tension between the expertise distributed across its workforce and the structures that determine whether that expertise can influence a decision. The agency’s Mission Management Team (MMT) had an opportunity to pursue evidence about damage to the space shuttle Columbia, but its hierarchical culture and tightly constrained discussions helped turn uncertainty into a settled conclusion. The disaster is therefore not treated simply as a failure to know; it is a failure to make what people knew, and what they still needed to learn, count collectively.

The Columbia damage assessment

On January 21, 2003, the MMT for NASA mission STS-107, Columbia’s twenty-eighth flight, held its second teleconference since launch on January 16. An hour before the meeting, Don McCormack had been briefed by the Debris Assessment Team (DAT), a group of engineers from NASA, Boeing, and Lockheed Martin. They had spent much of the preceding five days assessing the possible consequences of a large piece of foam that had broken off the left bipod area of the external fuel tank during ascent and struck Columbia’s left wing.

The available evidence was incomplete. No launch-tracking camera had captured a clear image of the impact, and a request for on-orbit pictures of the shuttle had not been approved. The DAT therefore estimated the foam’s size and impact speed, then used an algorithm called Crater to predict how deeply debris of that size and speed could penetrate the thermal-protection tiles on the wing. The team had reached no conclusion, but its members made clear to McCormack that the damage warranted concern.

That concern did not reach the MMT in the same form. The foam strike was not discussed until two-thirds of the way through the January 21 meeting, after topics including a jammed camera, the shuttle’s scientific experiments, and a leaky water separator. When MMT leader Linda Ham asked for an update, McCormack said that the possible damage and potential repairs were being investigated. He also noted that a similar strike on an earlier shuttle mission had caused fairly significant damage. Ham replied that she did not think the strike was much of a factor, since there was little that could be done about it. The discussion moved on. Before the team had reviewed the assessment, its leader had framed the issue as inconsequential for everyone in the meeting.

Hierarchy and the silencing of evidence

The book cautions against treating the MMT as the sole cause of the Columbia disaster: NASA had deeper institutional and cultural problems, many resembling those associated with the agency in 1986, when the Challenger exploded. Yet the investigation record also supports a narrower conclusion. The team had opportunities to make different choices that could have improved the crew’s chances of survival. Members were urged on multiple occasions to gather information needed to assess the shuttle’s safety, and they were warned that the foam might have caused burn-through: heat penetrating the protective tiles and reaching the fuselage during reentry. Even MMT leaders raised the possibility that the damage was serious. The team nevertheless failed to arrive at the right decision about what to do.

NASA’s self-image as meritocratic and bottom-up did not match the organization described in the book. The agency was deeply hierarchical, and the DAT engineers’ early concerns—especially their insistence that images of the Orbiter’s wing were needed for an informed analysis—never received a serious hearing from the MMT. The obstacle was not a lack of specialized knowledge. It was that the people closest to the technical uncertainty could not make their need for evidence decisive within the managerial discussion.

The team’s meeting practices compounded that hierarchy. Its discussions were both over-structured and under-structured: Ham generally posed a question and another person answered, but the team made no systematic effort to invite other members to address particular issues. A clear agenda and an active leader who ensures that everyone can contribute are associated with better group deliberation; the MMT’s format instead limited the range of analysis and information considered. On January 24, when McCormack summarized the DAT’s findings, he did not present the supporting analysis or explain whether the team’s members disagreed. No MMT member asked a question or requested the study. Ham’s brisk summary, “No burn-through, means no catastrophic damage,” effectively closed off debate.

Conclusions before evidence

The MMT began with a conclusion and interpreted subsequent information to fit it. That pattern made new evidence less influential even when people heard it: unfamiliar information could be discounted as unreliable or reshaped to match an existing picture of the situation. In the January 24 discussion, the absence of questions mattered as much as the summary itself. The team did not test how uncertain the analysis was, ask McCormack to quantify that uncertainty, or seek the underlying work. A meeting can appear to exchange information while leaving the shared assumption untouched.

The book relates this failure to research on small-group discussion. In an experiment by social psychologist Garold Stasser, groups assessing psychology students relied mainly on information shared by all members, even though the group collectively possessed useful details that only some participants knew. In unstructured conversations, commonly known information tends to be discussed more than information held by only one or a few people. Such evidence helps explain the MMT’s failure to make unfamiliar technical concerns consequential, without reducing the episode to a simple absence of information.

The missing diversity was cognitive rather than merely demographic. Former Mission Control operator James Oberg argued that the teams overseeing the Apollo missions had more diverse experience than the MMT: although the Apollo-era engineers might have looked alike, many had worked in different industries before joining NASA, while newer employees were more likely to have come directly from graduate school. Distinct professional histories can produce different interpretations of the same evidence. Research on mock juries by Chandra Nemeth likewise found that even a minority viewpoint can make group decisions more nuanced and deliberation more rigorous, including when that view ultimately proves mistaken. A dissenting MMT member pressing the possibility of serious wing damage might have forced the group to examine its conclusion more carefully.

Distributed knowledge, weak aggregation

The case echoes the book’s broader distinction between decentralization and aggregation. Decentralization encourages specialization and lets people draw on local, experience-based knowledge; its weakness is that valuable discoveries in one part of a system may not travel to the rest. Organizations need ways to make individual knowledge collectively useful without erasing the local expertise that produced it. The intelligence community’s pre-September 11 failures are described in similar terms: agencies held specialized information, but lacked a mechanism for aggregating judgments. Centralizing authority was not, by itself, the answer; the essential need was to connect and aggregate what different parts of the system knew.

For NASA, the gap between engineering analysis and managerial decision was not solved by simply having experts present. The DAT could estimate the strike and identify uncertainty, but the MMT’s hierarchy, meeting structure, and initial assumptions kept that analysis from changing the team’s course. In this account, better collective judgment depends on more than access to expertise: it requires channels through which local knowledge can be presented, questioned, and weighed, and a discussion that gives minority concerns room to alter the conclusion.