• Choose a category
• All
• Classic-Fiction

By Root 385 0

that criterion to understand why scientific problems are especially well suited for attack by collective intelligence. To understand the criterion, let’s first turn to an experiment done in 1985 by the psychologists Garold Stasser and William Titus. What Stasser and Titus showed is that groups discussing a certain type of problem—a political decision—often do surprisingly badly at using all the information they possess. This perhaps doesn’t sound so surprising: after all, everyday political discussion isn’t always terribly informative. But what Stasser and Titus showed went much further: group discussion sometimes actively makes people’s political decisions worse than they would have been if they had made those decisions individually.

Stasser and Titus began by creating written profiles of three fictional candidates for president of the student government at Miami University, where Stasser was a faculty member. The profiles contained information about the candidates’ policies on issues of interest to students—dorm visitation hours, local drinking ordinances, and so on. Stasser and Titus deliberately constructed the three profiles so that one of the candidates was clearly more desirable than the other two. They did this by first surveying students to figure out which traits students found desirable, and then constructing the profiles accordingly. We’ll give this extra-desirable candidate a name: we’ll call them “Best.”

In the first version of the experiment, each student received complete profiles of all three candidates, and was asked to decide who their preferrcandidate was. Not surprisingly, 67 percent of the students chose Best. Stasser and Titus then divided the students into small groups of four people each, and asked the groups to discuss which candidate should be president. At the end of the discussion the students were again asked for their preferred candidate. Support for Best increased to 85 percent.

So far, no surprises. But Stasser and Titus also did a second version of the experiment. This time they altered the profiles so that each student received only partial information about the three candidates: they removed some of the positive information about Best—things students could be expected to like—and they also removed some of the negative information about one of the undesirable candidates. In fact, any single partial profile now suggested that one of the undesirable candidates was actually better than Best. Not surprisingly, when asked to choose a candidate on the basis of these partial profiles, 61 percent of the students preferred the undesirable candidate, while only 25 percent preferred Best. After this, Stasser and Titus again divided the students into small groups of four, and asked the groups to discuss which candidate should be president. But here’s the clever bit: when Stasser and Titus were constructing the partial profiles, they were careful to remove different information from different profiles, so that each group of students would still have all the information about all three candidates. Thus each group still had all the information they needed to identify Best as the truly best candidate. Note that the students were warned in advance that not everyone in their group necessarily had the same information about all three candidates.

Now, in this second version of the experiment you’d think Best’s percentage would increase after the group discussion, as people shared what they knew and realized that Best was truly the better candidate. But that’s not what happened. In fact, after the discussion it was the undesirable candidate whose percentage increased, from 61 percent to 75 percent. Best’s percentage actually decreased, from 25 percent to 20 percent. The groups weren’t so much sharing information as they were reinforcing the students’ preconceived ideas. To put it another way, group discussion didn’t make the groups’ decisions better, it made them worse. It was a case of collective stupidity, not collective intelligence.

Whats was going on? We’ve seen many examples showing how groups can