U.S. Department of Defense

The Department of Defense is presented as both a powerful institutional center within the U.S. intelligence system and the sponsor of an unconventional effort to improve its forecasting. The contrast captures the book’s argument about organization: the challenge is not simply to centralize or decentralize, but to connect specialized local knowledge so it can inform collective decisions.

Position in the Intelligence System

Before the creation of a peacetime intelligence agency, the United States divided intelligence-gathering responsibilities among the military services. After the 1947 National Security Act created the CIA, that arrangement did not become a unified system. Instead, intelligence responsibilities spread across agencies with overlapping missions, including the National Security Agency, the National Imagery and Mapping Agency, the National Reconnaissance Office, the Defense Intelligence Agency, and the intelligence arms of the military services.

The CIA director was nominally responsible for the intelligence community as a whole, but exercised little practical supervision; much of the money for intelligence operations came from the Department of Defense. The resulting system consisted of largely autonomous groups pursuing the broad common aim of national security through different methods. The Department’s financial weight therefore coexisted with fragmentation rather than ensuring that information and analysis were brought together.

Fragmentation and the Case for Aggregation

The book distinguishes useful decentralization from the intelligence community’s particular failure to coordinate. Separate agencies can develop specialized expertise and independent perspectives: information that one set of analysts treats as irrelevant may reveal a clue or relationship when another examines it. But specialization is of limited value if findings and judgments do not travel between groups.

After the September 11 attacks, Congress emphasized the lack of information-sharing among agencies. Senator Richard Shelby criticized the FBI’s decentralized structure for leaving information holdings divided among largely independent fiefdoms; the broader intelligence community, too, had failed to put relevant information before the right people. Yet simply placing a central authority above the agencies would risk sacrificing the diversity, local knowledge, and independence that decentralization can provide. The book’s proposed remedy is aggregation: preserve specialized agencies while enabling their information and judgments to contribute to a collective assessment.

FutureMAP and Prediction Markets

The Department of Defense’s Defense Advanced Research Projects Agency funded FutureMAP, an abortive proposal to apply decision markets to intelligence. One component would have created small internal markets, perhaps involving twenty or thirty participants, open to intelligence analysts from different agencies and potentially a limited number of outside experts. Participants could buy and sell futures contracts based on expectations about events in the Middle East and elsewhere, drawing, where appropriate, on classified information and intelligence data.

The proposed markets were intended to offer an alternative to bureaucratic competition and pressure to produce conclusions acceptable to an agency or superior. By rewarding accurate forecasts, they might make the quality of a prediction more important than pleasing a boss or satisfying the White House, while drawing on analysts whose knowledge was scattered across agencies. The scheme thus treated market prices as one possible means of aggregating dispersed judgments without requiring a traditional top-down intelligence structure.

The plan also included the Policy Analysis Market, which attracted intense political criticism in the summer of 2003. Though public attention centered on disturbing possibilities such as wagers on an assassination, the book stresses that most proposed wagers would have concerned less sensational questions: Jordan’s future economic growth or the strength of Syria’s military. The market was meant to help assess the economic health, civil stability, and military readiness of Middle Eastern states and their implications for U.S. interests, rather than to predict a specific imminent attack.

Limits and Significance

FutureMAP’s logic did not make prediction markets infallible. If a market forecast a coup in Jordan and the Department of Defense intervened to prevent it, the intervention would make the forecast appear false and could undermine incentives for accurate predictions. A design might need to make wagers conditional on U.S. action, or traders might incorporate the prospect of intervention into prices. A sufficiently liquid market could also face familiar risks, including bubbles and gaming. The proposal’s significance, in the book’s account, lies not in a guarantee of perfect forecasts but in treating market aggregation as a potentially useful tool despite its limitations.

Across these examples, the Department of Defense embodies the tension between institutional power and collective intelligence. It supplied much of the intelligence system’s funding, yet financial centrality did not itself overcome the fragmentation of information. Through FutureMAP, the Department also backed an experiment in connecting separate judgments by a different mechanism: not by making one authority the sole interpreter, but by seeking a shared signal from many participants.