Cass Sunstein, Professor at Harvard Law School, is speaking today on Extremism: Politics and Law. Related to this topic, he is the author of Nudge, Republic.com 2.0, and Infotopia. He discussed Republic 2.0 with Henry Farrell on this bloggingheads.tv diavlog, which touches on the theme of extremism in discourse and the web’s role is facilitating polarization of political views (notably, Farrell gives a good counterfactual to Sunstein’s claims, and Sunstein ends up agreeing with him).
Sunstein is in the midst of writing a new book on extremism and this talk is a teaser. He gives us a quote from Churchill: “Fanatics are people who can’t change their minds and will not change the subject.” Political scientist Hardin says he agrees with the first clause epistemologically but the second clause is wrong because they *cannot* change the subject. Sunstein says extremism in multiple domains (The Whitehouse, company boards, unions) results from group polarization.
He thinks the concept of group polization should replace the notion of group think in all fields. Group Polarization involves both information exchange and reputation. His thesis is that like-minded people talking with other like-minded people tend to move to more extreme positions upon disucssion – partly because of the new information and partly because of the pressure from peer viewpoints.
His empirical work on this began with his Colorado study. He and his coauthors recorded the private views on 3 issues (climate change, same sex marriage and race conscious affirmative action) for citizens in Boulder and for citizens in Colorado Springs. Boulder is liberal so they screened people to ensure liberalness: if they liked Cheney they were excused from the test. They asked the same Cheney question in Colorado Springs and if they didn’t like him they were excused. Then he interviewed them to determine their private view after deliberation, and well as having come to a group consensus.
Sunstein found that views they liked turned into views they loved and vice versa after discussion with the like-minded. This is a shift in *anonymous views*. And also, the internal diversity in the groups’ views that existed before discussion was squelched after they meet.
Sunstein extended this by examining voting patterns of 3 judge judicial panels. Do Democratic appointees vote differently if they are on panels with all dems or with mixed panels? And Republican appointees? Depending on the subject of the case it appears the same extremism appears when a judge is surrounded by other judges appointed by a president of the same party.
He examined about 35k votes (judges are randomly assigned to panels) and, for example, in gay rights cases from 1996-2006 a Democrat appointee on DDD panels rules against gay rights 0% of the time. A Republican appointee on RRR panels vote against gay rights 86% of the time. In sex discrim cases: D on DDD panels vote for the plaintiff about 72% of the time, R on RRR panels about 25% of the time. This is a typical pattern which he analogizes to what happened in Colorado.
So for example, a Republican appointed judge on a mixed panel with be liberal about 38% of the time. When he or she in on an RRR panel the vote drops to 24%. A democratic appointed judge is liberal about 50% of the time of mixed panels, and on DDD panels this jumpe to 62%.
There are three areas where this effect is not observed: capital punishment, abortion, and national security after 9/11. Here the judges seemed impervious to panel composition.
The third study he outlines is on that gets at rhetorical bias. He looked at outrage in the case of punitive sentencing. About a 1000 people were asked to rate whether the action was outrageous and whether the award was outrageous. Generally people agreed on the action (a kids pyjamas catching fire was outrageous, but a failed balding remedy was not outrageous) but differed over the outrageousness of the financial compensation.
They created millions of statistical juries of 6 from this sample, then they predicted the jury outcome by taking the median of the individual responses. This appears to predict the level of outrageousness well but not the dollar award where the predictions are erratic.
But is the median a reasonable way to predict? They made a mock jury study to test this: it turns out the median is not a good predictor. When people are upset, the outcome went above the median (they began with antecedent outrage), when the outrage is low then th eoutcome shifted in a lenient way. With dollars this is more dramatic – dollar awards almost always award higher than the median. 27% of the cases jury was at least as high as the highest juror!
Sunstein thinks the problem with his jury study is that it doesnt take group polarization into account – deliberation makes groups more extreme. This is discovered at MIT in the 1970’s: a business student got people to deliberate on various entrepreneurial deicsions. Stoner found that these students were more risk inclined than the median member was before they started to talk. But there were two issues upon which the MIT grad students showed a risk averse streak: boarding a plane when you’re sick and whether you should get married when you’re not sure. A subsequent study did this with citizens of Taiwan: the Taiwanese got more cautious where the Americans got more risk loving. Sunstein thinks this applies to fed judiciary study.
So what accounts for this shift towards extremism? Sunstein has three ideas:
1) exchange of information: a like-minded group won’t discuss where they agree so this amplifies their concern. For example, environmentalists won’t discuss that climate change isn’t a problem but that it is. So they get more alarmed.
2) reputation factor: you want to portray yourself as more extreme than the median, so when you are in a more extreme group you might shift to keep your position as more extreme than the median. We also tend to judge other people as more favorable if they agree with us. But interestingly, when this agreement is found the increase in favorability also works for the other person.
3) When people don’t know what they think they tend to go for the middle. When people find their inital views are corroborated they hold their initial view more strongly and tend to go more extreme. Sunstein thinks this is happening everyday on the internet – people can easily corroborate their inital inclidinations.
But there’s a wrinkle: why are the post deliberation dollar awards higher than the median or even the highest juror? Kahnemen thought about this and postulated that there is a built in rhetorical assymetry between those seeking higher vs lower damage awards. Sunstein tested whether those favoring higher awards have a rhetorical advantage: He asked Chicago Law School students how they would argue for a higher award if they were on a jury with a punative damage award (without knowing anythign about the case). And he asked another group how they woudl argue for a lower award. Then he asked both groups which direction was easier to argue for. Both groups said it was easier to argue for the higher award. Arguments for the higher award were deterrance and compensation (society-wide effects), and all the other side had was that this would be a windfall for the plaintiff (at worst mistake for a single person who is not made worse off). This applies to tax increases, criminal punishment…
Sunstein notes Madison’s solution to extemism: mix it up! Mixed panels are superior to RRR and DDD panels..