Two heads sometimes better than one

Wise groups follow the advice of confident decision makers. So do foolish groups. But a group’s success or failure depends on whether what’s commonly known corresponds to the truth in a given situation, a new study finds.

Heeding recommendations of the most confident member of a two- or three-person group often works well, says psychologist Asher Koriat of the University of Haifa in Israel. That’s because in many settings, high confidence is associated with majority opinions, which are frequently correct about general types of knowledge, Koriat reports in the April 20 Science.

Confidence-based group decisions go awry when majority opinions don’t jibe with reality. Most people assume, for example, that the larger of two countries in area also has a larger population, but there are exceptions to this rule. In these instances, the opinion of the least confident group member is most accurate, Koriat observes.

This finding raises the provocative possibility that, in uncertain environments where common knowledge can’t be trusted or doesn’t exist, groups should rely on the guidance of those who express the most doubt about a decision.

Opting for hesitant over self-assured suggestions when the going gets tough doesn’t come easy. “Can you think of a group that would be willing to follow the lead of the least confident member?” Koriat asks. “I cannot.”

Koriat assembled data from individuals who had completed experimental tasks into simulated two- and three-person groups. Group decisions thus involved no negotiations or haggling. Instead, accuracy rates were calculated for simply going with the choices of more- or less-confident group members on each lab problem.

Koriat’s findings challenge studies suggesting that overconfidence typically leads decision makers astray. That’s because experimenters often cherry-pick questions so that what people assume to be true based on their experience yields incorrect answers, remarks psychologist Ralph Hertwig of the University of Basel in Switzerland.

One such example: It’s relatively easy to infer that more people live in Paris than in Toulouse. Paris is widely known as the capital of France, and in many countries the capital is the largest city. Having confidence that capitals are more populated than other cities makes sense. Yet that handy rule of thumb sometimes doesn’t work. In Switzerland, for example, residents of Zurich outnumber those of Bern, the nation’s capital.

Koriat examined how well a confidence-based strategy works in groups. He administered general knowledge questions, involving choices between two options, to individual college students. In some trials, volunteers who expressed a lot of confidence in their choices were usually right, suggesting they had tapped into a store of common knowledge. In other trials, high-confidence choices were frequently wrong.

Participants rated their confidence in each of their answers. Koriat then created simulated groups by analyzing responses of pairs of volunteers with different accuracy rates and confidence levels. When most people knew the correct answers to problems, two-person decisions based on the more confident member’s choice were more accurate even than those of the best-performing individual. In another test, Koriat showed that three-person decisions guided by the most confident member slightly outperformed two-person decisions.

When most people didn’t know the answers, confidence-based group decisions were worse than those of even the worst-performing individual.

“One ought to be conservative in deciding whether to trust group decisions, because the price might be heavy if the issue is one for which people are predominantly incorrect,” Koriat says.