There is a general consensus that many things are wrong with the way U.S. intelligence agencies operate. The weaknesses of the present system have been clearly brought to light by the 9/11 commission’s report. Included in that report is a recommendation to create the new position of national intelligence director to oversee the15-agency intelligence community, which includes the CIA, the FBI, the NSA and the armed forces intelligence organizations. The commissioners said that without a more integrated approach to intelligence, “it is not possible to 'connect the dots’ “ about terrorist intentions.
An important question is: Will the appointment of a national intelligence director really make it possible to better connect the dots? I don’t believe it will. Recent research that I conducted about consensus building, combined with earlier work by Professor Serge Galam of France on hierarchical organizations, suggests that a solution to more effective communication and intelligence sharing lies elsewhere.
The public discussion on this issue has so far focused on the perceived dangers of granting such great power to a single individual, especially one who is not elected. Such fears are not unfounded. History clearly shows the dangers of similar concentrations of power. During his 50 years at the helm of the FBI, J. Edgar Hoover was able to influence the decisions of elected officers through the secret files he collected on many private and public citizens.
Surprisingly, no discussion has taken place about the assumption that a adding another “layer” at the top of the intelligence community would actually improve the likelihood of being able to “connect the dots” about terrorist intentions. This lack of discussion is based on the sense held by most of us that the best way to make decisions is to have a leader at the top who collects information and makes final decisions based on that information. This interpretation bias is evident in many situations. When asked why migrating geese fly in “V” formations, most of us answer that the geese are flying behind the leader of the flock. In reality, the goose now at the tip of the formation is there only temporarily. The flock has no leader.
The problem with hierarchical structures is that communication occurs mainly along vertical lines. Unfortunately, this characteristic of these organizations permits the emergence of “consensus” at the top levels that are actually in disagreement with the majority assessment of the situation. Imagine a hierarchically-organized intelligence agency that is investigating if country “X” holds weapons of mass destruction (WMD). Imagine also that the individuals at the higher levels of the agency have a bias toward a positive answer to the question. Then it is quite possible to have a situation in which more than 75 percent of individuals in the organization think that there are no WMD but in which the “consensus” view of the organization is that country “X” possesses WMD.
The situation is, however, not without hope. Humans have developed many processes by which to share information globally without the need for a central authority to collect that information or without having to meet face-to-face with everyone else. In a recent study published in the Proceedings of the National Academy of Sciences, a team I led demonstrated that it is possible to devise robust and efficient communication strategies by which information can be shared and a consensus reached without the need for a central authority or a hierarchy. Remarkably, our study demonstrated that such communication strategies are especially efficient in the presence of noise, that is, when some of the individuals do not share information truthfully.
The results of our study merely confirm that the principles at the core of democratic institutions provide a highly-efficient system to gather and process information in an uncertain and complex world. In the past, we came together in town hall meetings to share the information available to each of us so that a situation could be best assessed and a good solution found. Nowadays, there are too many of us for face-to-face meetings but the principle of sharing unadulterated, unfiltered information may still be implemented if one wants to make use of the many heads at the bottom instead of the single head at the top.
A decentralized solution to the sharing of intelligence has several advantages. First, there is no need for potentially dangerous concentrations of power and/or information. Second, the system does not break down if the person at the top is absent or untrustworthy. Third, there is no risk of influence by special interests or of obtaining information from a small and not representative set of individuals. Research shows that the solution to improve intelligence gathering and processing is not to add another layer at the top but to diversify channels of communication at the bottom.