Your independent source for Harvard news since 1898

Your independent source for Harvard news since 1898


Games of Trust and Betrayal

March-April 2006

“People care not only about outcomes, but about how outcomes came to be,” says associate professor of public policy Iris Bohnet of the Kennedy School of Government. “That doesn’t strike anyone but an economist—like me—as a surprise.” Game theory, as conceptualized by conventional economics, suggests that players care only about substantive results. With Ramsey professor of political economy Richard Zeckhauser, Bohnet developed a concept of “betrayal aversion,” building on the well-established psychological principle of risk aversion—by and large, humans simply don’t like to take risks.

It turns out they don’t like to trust, either, because trust is a form of risk that makes one vulnerable to betrayal. To buy an item on eBay, one must trust the seller. We also trust attorneys, doctors, and politicians to tell us the truth and to represent our interests. “These are principal-agent relationships,” Bohnet says. “An agent does something on your behalf. But principals’ and agents’ incentives are not always completely aligned, and there’s asymmetric information.”

Traditionally, academics have linked trust to risk tolerance, since it involves taking a risk. Instead, “We’re saying that risk-taking when the agent of uncertainty is nature is very different from when the agent is another person,” Bohnet asserts. A farmer, for example, faces natural risks like weather and soil conditions. But there are also social risks—speculative bubbles, HIV infection, terrorism—where other people produce the uncertainty.

Bohnet and Zeckhauser have been running two games, now with about a thousand subjects around the world, playing in groups of 30 at a time. They are two-person games, a variant of the classic Prisoner’s Dilemma. In the first game, Player A can choose a “safe” alternative or choose to trust Player B, who can in turn choose an option that rewards both of them more than the safe alternative, or a second option that brings even greater winnings for B—but less than the “safe” option would have given A. In other words, the “good” (i.e., trustworthy) B player will take the win-win alternative, while the “selfish” B will maximize his own outcome at A’s expense. When the researchers ask subjects playing A, “What percentage of good people would there have to be in the room [of 15 potential B players] before you would be willing to trust this stranger [the B player]?” the answer has consistently been 50 to 80 percent.

Associate professor of public policy Iris Bohnet, who has played games that measure “aversion to betrayal” with subjects from Brazil to Switzerland to Kuwait
Portrait by Stu Rosner
The second game has the same rules as the first, except that an urn containing 100 blue and green marbles takes the role of Player B. The urn is a proxy for an impersonal force, such as nature. If a blue ball is randomly chosen, B selects the “trustworthy” win-win alternative; if a green ball, the “selfish” one. The researchers then asked A players, “What percentage of blue balls would the urn have to contain for you to be willing to take this risk?” A rational money-maximizing person—one who cares only about outcomes—would give the same answer to this question as to the analogous one in the first game. But when playing with “nature,” respondents generally peg the figure at 30 to 40 percent, far lower than in the first game. “People are less willing to take risks when confronted with another person than when confronted by nature,” Bohnet explains. “Trust is not only about willingness to take risks, but about the willingness to be betrayed.”

By comparing the difference between “Minimal Acceptable Probabilities” in the first and second games, the researchers have been able to distinguish risk aversion from betrayal aversion. The “nature” game establishes a baseline level of risk aversion, but the game with a human Player B introduces the additional possibility of betrayal. Thus, the gap between percentages on the two games gives a rough index of betrayal aversion. In the United States, Switzerland, and Brazil, the betrayal aversion differential is 10 to 20 percent. Zeckhauser and Bohnet have also played the games in the Persian Gulf region, with subjects in Kuwait, Oman, and the United Arab Emirates. (They are the first social scientists to run economic experiments in the Gulf region, and will go to Saudi Arabia in March.) In these countries, betrayal aversion is markedly higher, with a differential in the 30 to 40 percent range. “Many in this area say they are willing to trust only if 100 percent of the people are trustworthy,” Bohnet reports.

She had an enlightening experience when teaching negotiation and decision analysis to a group of government ministers from the Persian Gulf region in a Kennedy School executive-education program. “I started the class by asking them to recall a time when they lost trust in someone,” Bohnet recalls. “One minister said, ‘Trust is not an issue for us. We never trust.’ What a beginning! It opened up a very interesting discussion. A minister said, ‘We cannot dare to trust because we may lose face. I would never come to a meeting and put something on the table that other people could decline.’ The meeting-before-the-meeting is absolutely critical in the Gulf, because being let down is terribly humiliating.”

Trust has other policy implications. Social capital, per capita income, economic growth, and political stability all have positive correlations with trust in a society. “Trust is a generally good thing,” says Bohnet. And nations deal with breakdowns of trust in different ways. “In the Western world, especially the United States, contract law builds on the notion of damages or efficient breach,” Bohnet says, “meaning that someone who breaches a contract must compensate their counterpart. But if people are really betrayal averse, damages won’t satisfy them, because what they are concerned with is the fact of betrayal. U.S. contract law focuses on decreasing the material cost of betrayal, but what betrayal aversion asks for is to decrease the likelihood of betrayal, which causes emotional hurt. In Islamic law, which seems to encourage building trust by personal relationships rather than legal means, damages play a much smaller role than in the West. In addition to differences in law, there obviously are other contributing factors. For example, group-based social organization, typical in the Gulf but not in Western countries, is based on long-standing relationships. This substantially reduces the likelihood of betrayal and thus, the social uncertainty involved in trust.”