Central Intelligence Agency of the United States

The Central Intelligence Agency was established in 1947 to bring unity to American intelligence after the failure to anticipate the attack on Pearl Harbor exposed the risks of fragmented responsibility. But the agency did not become the comprehensive command center its advocates had sought. The book’s central distinction is that decentralization can preserve expertise and local knowledge, while a lack of information sharing and judgment aggregation can leave a system unable to recognize the significance of what it already knows.

Creation and the case for unity

The agency’s origins lay in the division of intelligence-gathering responsibilities among the military services before World War II. None of those services anticipated the attack on Pearl Harbor, despite evidence that, in retrospect, appeared to point toward a major Japanese strike. The failure suggested that separate intelligence channels could leave the country vulnerable to surprise. The prospect of conflict with the Soviet Union and the arrival of technologies such as rockets, atomic weapons, and bacteriological warfare added urgency to the argument for a stronger peacetime intelligence organization.

In April 1946, “Wild Bill” Donovan made that argument publicly at a forum organized by the New York Herald-Tribune. Donovan had led the Office of Strategic Services, the principal American intelligence organization during World War II, and after the war became an outspoken advocate for a more powerful peacetime successor. In a speech titled “Our Foreign Policy Needs a Central Intelligence Agency,” he called for a centralized, impartial, independent body to take charge of the country’s intelligence operations. The proposed agency was meant to correct the weaknesses exposed by divided responsibility: unity of command and unity of intelligence would, its advocates hoped, make the country safer.

Donovan’s criticism of the intelligence community did little for his own career. It alienated people within that community and probably ended his prospects of returning to government service. Yet in 1947 Congress passed the National Security Act and created the Central Intelligence Agency. The creation answered Donovan’s call at the level of institutional design, but it did not produce the unified system that the law’s stated goal implied.

An agency without command of the whole

The Central Intelligence Agency initially occupied a central place in the postwar system, but the broader intelligence community became more fragmented over time. It developed into a network of agencies with overlapping responsibilities and missions: the National Security Agency, the National Imagery and Mapping Agency, the National Reconnaissance Office, the Defense Intelligence Agency, and the intelligence branches of the three major military services, alongside the CIA. Although the CIA director was theoretically in charge of the community as a whole, in practice the director exercised little supervision over the other agencies. Most intelligence funding came from the Department of Defense, further limiting the force of the formal claim to central authority.

The Federal Bureau of Investigation, responsible for domestic law enforcement, operated almost entirely outside the intelligence community’s orbit, even though information about foreign terrorists inside the United States was plainly relevant to the CIA. The result was not a single repository where intelligence could be gathered and analyzed as one coherent whole. It was a collection of largely autonomous groups pursuing the same broad goal—protecting the United States from attack through different institutional channels.

This arrangement complicates a simple verdict against decentralization. Separate agencies can specialize, interpret evidence through different analytical perspectives, and draw on local knowledge. A detail that appears irrelevant to one analyst may disclose a clue or relationship when considered by another. The structural weakness was not simply that intelligence work was divided. It was that the divisions were not matched by reliable means to share information and combine judgments. The system could accumulate many distinct observations without turning them into a collective assessment.

September 11 and the problem of missed connections

Before September 11, 2001, the shortcomings of the system had not forced a decisive reconsideration. American intelligence had failed to anticipate the 1993 bombing of the World Trade Center, the 1998 bombings of the U.S. embassy in Kenya and the USS Cole in Yemen, and then the attacks of September 11. After the attacks, a congressional inquiry concluded that the intelligence community had failed to make use of the individual and collective significance of information that appeared relevant. It also found that agencies had missed opportunities to disrupt the plot and had let information pass unnoticed that, if recognized, could have improved the chances of discovering and preventing it.

The comparison with Pearl Harbor was powerful: a familiar pattern of surprise after information had existed in scattered form. But the account also cautions against treating hindsight as proof that the attacks should have been predicted. A retrospective inquiry can locate clues that seem meaningful once the outcome is known; it does not follow that analysts could reliably have identified them in advance. Roberta Wohlstetter’s history of the Pearl Harbor failure emphasizes the difficulty of selecting useful signals from a mass of confusing information. Strategic surprise is hard to overcome, and the challenge of anticipating an attack by nineteen terrorists was different in scale from foreseeing a large Japanese naval assault.

The question, then, was not whether a particular attack could certainly have been predicted. It was whether a different organization might have improved the chance of recognizing that the available information pointed toward a serious threat. Congress’s answer emphasized process: better procedures could yield better results. The Joint Inquiry identified failures of information sharing, in which separate agencies produced localized snapshots instead of a coherent picture. Senator Richard Shelby argued that the FBI’s decentralized structure had left information holdings in largely independent fiefdoms. He urged a return to Donovan’s idea of an agency positioned above and independent of the competing bureaucracies.

From centralization to aggregation

The proposed cure of a commanding central agency, however, risked sacrificing the very advantages that specialization and decentralization provide. A single center might make information easier to collect under one authority, but it would not necessarily make a small group of decision-makers wiser than the larger community. The account’s alternative is to distinguish the act of centralizing control from the act of aggregating knowledge. The intelligence community needed information to be shared, but that did not mean one person or a small central team should do all the sifting and judging. Its collective analytical capacity was itself a resource; the task was to create mechanisms that could bring that capacity to bear.

Some remedies were modest and technical. Linking the agencies’ computer databases could have improved information flow without dissolving their autonomy. Yet two years after September 11, the government still lacked a single unified watch list drawing on data from across the intelligence community. The continuing absence of such a basic connection showed how organizational fragmentation could persist even where a comparatively mechanical fix was available.

A more ambitious proposal was FutureMAP, an abortive program funded by the Defense Advanced Research Projects Agency. It would have used decision markets, similar to the Iowa Electronic Markets, to let analysts from different agencies and bureaucracies buy and sell futures contracts tied to expectations about events in the Middle East and elsewhere. One proposed component consisted of small internal markets, perhaps involving twenty or thirty participants, open to intelligence analysts and possibly a limited number of outside experts. Rather than placing all interpretive authority in a single hierarchy, such a mechanism would have offered a way to aggregate dispersed judgments. The proposal did not establish a new intelligence center; its significance lay in the attempt to connect specialized assessments without erasing the institutions that produced them.

The broader organizational principle is that decentralization and integration are not opposites that must be resolved by choosing one. People close to a problem may possess tacit knowledge that is difficult to summarize or transmit, and local specialization can enlarge the total amount of information available to a group. But local knowledge contributes to collective intelligence only if the system can carry it beyond the place where it originated and combine it with other people’s judgments. The CIA’s history, in this account, therefore turns on an unfinished institutional problem: how to preserve diverse, independent expertise while making its separate findings collectively useful.