Cooperation problems

Cooperation problems are situations in which self-interested, distrustful people must work together even when narrow individual incentives appear to favor opting out. The book treats cooperation not as an automatic consequence of shared interests, but as an achievement shaped by reciprocity, rules, incentives, and the ways a group shares and combines information.

The Problem of Working Together

The central difficulty is the gap between individual advantage and collective benefit. Paying taxes, addressing pollution, and agreeing on reasonable pay all require people to accept costs or limits that may not seem worthwhile if considered in isolation. Yet when participants withhold their contribution, the shared undertaking can fail, leaving the group worse off than if its members had cooperated.

Cooperation is one of three broad kinds of collective problem considered in the book, alongside cognition and coordination. Cognition asks for answers that can be judged against a solution or against better and worse alternatives; coordination asks people to align their behavior while anticipating one another. Cooperation instead concerns whether people who have reasons to distrust one another will contribute to a common undertaking at all. The book’s structure moves from these theoretical problems and the conditions that support collective intelligence toward case studies of how people organize around shared, or only partly shared, goals.

Incentives, Reciprocity, and Free Riding

Public-goods experiments conducted by economists Ernst Fehr and Simon Gächter make the conflict tangible. In each group of four, participants began with twenty tokens and played four rounds. A contributed token cost its owner one token, but returned 0.4 tokens to each of the four players: a total group gain of 1.6. If everyone kept their tokens, each finished with twenty; if everyone contributed all of them, each finished with thirty-two. For an individual, however, the tempting strategy was to keep personal tokens while benefiting from everyone else’s contributions. If all followed that strategy, the public pot would receive nothing.

Participants generally did not begin by behaving as pure free riders. At first, most contributed about half their tokens. But as the rounds went on and they saw others withholding contributions, giving declined. By the end, 70 to 80 percent of players were free riding, and the group as a whole was poorer than it would have been under sustained contribution. The sequence shows how cooperation can erode through reciprocal disappointment: observing others’ defection changes what seems reasonable for each person to do next.

Tax compliance follows a related logic. In the United States, people are willing to pay what they regard as their fair share when they believe others are paying too and that tax evaders are likely to be caught and punished. Historian Margaret Levi calls such taxpayers “contingent consenters.” Their willingness depends on reciprocity and credible enforcement; if they think others are breaking the rules without consequence, they feel taken advantage of. Cooperation, in this account, rests neither on self-interest alone nor on an assumption of universal goodwill, but on beliefs about other people’s conduct and the rules that govern it.

Rules, Information, and Collective Judgment

The book’s account of cooperation connects it to the broader design of collective decisions. In classroom experiments on information cascades, students who chose sequentially often followed earlier guesses rather than relying on their own private evidence. Angela Hung and Charles Plott changed the incentives: instead of paying students for an individually correct answer, they rewarded them according to whether the group’s majority answer was correct. Students then had reason to attend more closely to their private information, even though it could lead to a wrong individual guess. The group’s collective judgment became significantly more accurate than that of cascading groups. The experiment illustrates how changing what participants are rewarded for can encourage behavior that serves the group rather than merely the individual.

That principle also shapes the book’s discussion of intelligence work. Before September 11, the intelligence community had specialized agencies and diverse local knowledge, but lacked a reliable means of aggregating judgments across them. Richard Shelby’s proposed remedy, a truly central intelligence agency, would have addressed the organizational gap by putting information under central authority, but risked sacrificing the independence and diversity that decentralization makes possible. The book distinguishes centralization from aggregation: information must travel and judgments must be combined, but a small group at the center need not be the only way to accomplish this.

One proposed alternative was FutureMAP, an abortive plan funded by the Defense Advanced Research Projects Agency. Its internal markets would have allowed intelligence analysts, and perhaps a few outside experts, to trade on forecasts of events, potentially drawing on classified information. By rewarding accurate forecasts rather than conformity to a supervisor, agency, or White House, the markets were intended to counter bureaucratic pressures and aggregate judgments across agencies. The proposal’s policy-market component became a target of public criticism in the summer of 2003. In the book’s treatment, the case raises a recurring design question: how can institutions make scattered knowledge useful to the group without extinguishing the independence that produced it?

Decentralization has a corresponding weakness. It can encourage specialization and preserve knowledge grounded in a particular place or job, but information discovered in one part of a system may never reach the rest. Cooperation therefore requires more than local autonomy: a group must also find ways to make local knowledge collectively useful. The book presents this balance as a practical condition for solving difficult problems, not as a choice between unrestricted independence and a single commanding center.

Cooperation in Science

The international search for the cause of SARS offers a contrasting case in which collaboration converted dispersed work into a collective result. After reports of a severe respiratory disease emerged, the World Health Organization brought eleven laboratories from countries around the world into a joint research effort. Researchers held daily teleconferences, posted microscope images, analyses, and test results on a WHO website, and exchanged virus samples. Because laboratories worked in parallel and could check one another’s findings, they narrowed the possible causes quickly. By March 21, researchers in Hong Kong and at the Centers for Disease Control had separately isolated promising candidates; subsequent testing identified a coronavirus in many SARS samples, and infected monkeys in the Netherlands developed the disease. On April 16, about a month after the collaboration began, the laboratories were confident that the coronavirus caused SARS.

The discovery could not be credited to a single person. Cynthia Goldsmith first spotted the coronavirus, but establishing that it caused SARS required the work of laboratories around the world, including the elimination of other possible causes. The WHO’s account described the group as having collectively discovered the virus. Nor was the network run by a director assigning every task: the WHO convened it, while the laboratories agreed to share relevant data and speak each morning, then worked out how to divide the investigation. Here cooperation made specialization productive because participants shared evidence and could build on one another’s findings.

The episode exemplifies a wider scientific norm: publish results and make data available so that other researchers can inspect, challenge, and use them. Scientific knowledge depends on a common pool that is continually replenished; even failed hypotheses can help by showing other researchers where not to look. The SARS case gives that principle an operational form: shared samples and daily exchange allowed independent teams to test the same question at once, rather than leaving each to work alone for months or years.

Trust, Standards, and Organizational Failure

Scientific cooperation also depends on trust that does not eliminate scrutiny. Once the laboratories’ evidence was examined and accepted by the scientific community, researchers could proceed on the basis that SARS was caused by a coronavirus. Most scientists do not personally replicate every experiment on which they rely; they trust that reported data and methods are sound, while replication and the possibility of exposing fraud remain part of the system. The book emphasizes that collective knowledge therefore depends both on communication and on confidence in the community’s ability to distinguish credible findings from unreliable ones.

Organizations can undermine that confidence when their incentives punish candor or reward misleading reports. Research by Chris Argyris describes inauthentic behavior as common in organizations, where hostility from bosses toward opposition can prevent information from moving upward. Performance targets create another distortion. In sociologist Donald Roy’s early-1950s machine-shop study, lathe operators were paid under a piece-rate system with successive output thresholds. If workers exceeded the hurdles too readily, management might raise them; the workers consequently restricted output and concentrated on manipulating the rate rather than maximizing production. Roy called the practice “Goldbricking in a Machine Shop.”

The same incentive problem can reach management. When bonuses depend on surpassing targets, managers have reasons to set undemanding expectations and then to use accounting tactics to meet them, even at the expense of future results. In such systems, the organization formally asks for performance but rewards information and behavior that make honest planning harder. Across the book’s examples, cooperation depends on more than wanting a common outcome: rules must make contribution credible, information must be able to travel, and incentives must not turn candor or effort into liabilities.