In the roughly three decades between the election of Ronald Reagan and last autumn's global financial collapse, social scientists and public-policy thinkers were obsessed with the way society seemed to grow more fragmented as it grew more prosperous. Almost everyone diagnosed a growing gap between rich and poor, but the polarization was not just economic. Mickey Kaus showed that people were mingling less in public spaces. Richard Herrnstein and Charles Murray showed that people were sorting themselves by measured intelligence and scholastic achievement.
It was Cass Sunstein, now a Harvard constitutional law professor, who first alerted a broad public to the kind of polarization that has preoccupied us most in recent years. Society, with the help of the Web, was sorting people by ideology in a way that eroded fellow-feeling and fostered mindless partisanship. Almost a decade ago, his Republic.com lamented that while daily newspapers confront people with all kinds of material they didn't ask for, the Web allows them to dodge what they disagree with. This was an alarming refutation of our smug claims about the Internet. In theory, the Internet opens people up to new ways of looking at things. In practice, it lets people wall themselves off in informational micro-environments of their own design. It makes them not more cosmopolitan but more parochial.
Now Sunstein has written Going to Extremes, a short book about the nature and roots of extremism. It is meant to unsettle us in the way his earlier work did. He finds that sitting people down to deliberate does not necessarily lead them to compromise or to converge on their mean opinion. They tend to radicalize in the direction of whatever bias they had to begin with. Teams of doctors, deciding collectively, are more likely to support the "extreme" strategy of heroic efforts to save terminally ill patents than the average individual doctor among them. Juries tend to vote, after discussion, for much more "extreme" monetary awards than the average individual juror among them would. Talking things over isn't necessarily wrong. But it doesn't lead reliably to moderation, either.
Other people have made similar arguments. To take an example that Sunstein does not mention: When Barack Obama won the Democratic Party's nomination last year, largely thanks to his strength in caucus states, Hillary Clinton's supporters complained that the deliberative caucus system didn't just express voter sentiment but warped it. It would be interesting to know whether Sunstein—President Obama's friend, former colleague, and nominee as chief White House regulator—agrees with the Clinton view. He seems to.
Much of Sunstein's evidence about how people drift to extremes comes from his studies of groups that already have a bias to begin with. Individual Democrats and Republicans on three-judge panels cast more "extreme" votes when they are in the majority than when they are not. A group of conservative Republicans in Colorado Springs will move sharply rightward when they discuss global warming among themselves, and a group of liberal Democrats from Boulder will move sharply leftward.
These homogeneous groups are not the special cases they would appear. They tell us something about what happens in more heterogeneous groups, too. If you bring the two clashing sides together, they don't find middle ground any more than like-minded people do. Each side digs in. If you give "a set of balanced, substantive readings" to a group that is at loggerheads over abortion or affirmative action, Sunstein shows, each side simply mines the readings for support of its own position. Ideology, it turns out, is not just a matter of opinions or positions—it is a predisposition to receive some kinds of evidence and not others. Compounding the problem, certain kinds of extremist arguments have an "automatic rhetorical advantage" in deliberation. Me, too, but less is harder to rally behind than In for a penny, in for a pound. In recent years, there have been no effective arguments against endlessly ratcheting up drug sentences or endlessly ratcheting down tax rates.
These insights—which come mostly from the first 30 or so pages of a 150-page book—are immensely interesting, but Sunstein has a hard time building anything out of them. To be fair, it is not clear that he means to. The book is less a fully mustered argument than a collection of write-ups of half a dozen research projects clustered around the social dynamics of extremism.