Thursday, August 12, 2004

Why Group Think ?

The major error on Iraqi WMD has been blamed on group think -- the tendency for people to arrive at a consensus and for the opinions of people who disagree to be supressed often by themselves. Why do people do this ?

First it is overwhelmingly clear that people do exactly this. The experiments concern simple questions about which no semi normal person can have any doubt such as which of two lines on a black board is longer. When all but one person in a group invited to answer such a question are shills and give the wrong answer the one subject often goes along (I don't remember the cite on this). This might be a feature of how people act in weird situations in which everyone else must be crazy or lying. More generally, the opinion expressed by a group after discussion tends to be more extreme than the average opinion before discussion. This might be due to rational pooling of information. Together there is a strong case for group think. Another name for group think is herd behavior.

So why is there group think ? There are three simple explanations.

First people might systematically underestimate the quality of their personal information and over estimate the quality of the information of others in the group. This means that when arriving at a post discussion opinion they give too little weight to their own pre-discussion opinion. This seems unlikely, since there is strong evidence that people overestimate the quality of their personal predictions

Second if people are rewarded based on their forecasting performance compared to average forecasting performance and are risk averse, they will rationally state a forecast closer to the average than their sincere subjective mean squared error minimizing forecast.

Finally herd behavior might be, well, behavioral. That is we might have an inate inclination to agree with what other people are saying. This would make sense, since a group that is divided by different opinions risks splitting or being paralyzed.

I suspect the second and third explanation are both important. I think it is possible to decide which matters. In experimental settings, it is possible to change peoples incentives. For example if the prize given to the subjects is the same for each subject and based on the difference between the average forecast and the outcome herding would be irrational.

In contrast, as noted by Canice Pendergast, if relative performance valuation is important, one should see a contrast between apparently impetuous youths and sage old timers. The point is that if someone has a good reputation, they risk losing it by being more wrong than anyone else. If someone has no reputation, they have a chance to win big by being more right than anyone else. This theory fits the fact that the main stream of opinion is very narrow compared to, say, opinions you find on blogs.

For example when explaining the Washington Posts feeble resistence to Bush administration Spin on Iraqi WMD Howard Kurtz discusses the issue with Bob Woodward (who has high standing)

"Woodward, for his part, said it was risky for journalists to write anything that might look silly if weapons were ultimately found in Iraq. Alluding to the finding of the Sept. 11 commission of a "groupthink" among intelligence officials, Woodward said of the weapons coverage: "I think I was part of the groupthink." "

No comments:

Post a Comment