One of the most interesting findings in modern social science involves group polarization -- the process by which like-minded people go to extremes. More technically, deliberating groups tend to end up in a more extreme position in line with their predeliberation tendencies. (A good discussion can be found in Roger Brown, Social Psychology: The Second Edition (1985); at the University of Chicago the phenomenon has been explored for both juries and three-judge panels.) It follows, for example, that after talking with one another, those who are excited about Judge Alito will be still more excited about him; those who are skeptical of him will be still more skeptical after internal discussions. (Of course this is a statistical regularity, not an unbroken rule.)
It is useful to distinguish between two different kinds of polarization: planned and spontaneous. Some people act as "polarization entrepreneurs": they attempt to create communities of like-minded people, and they are aware that these communities will not only harden positions but also move them to a more extreme point. Some of history's greatest heroes and villains operated as polarization entrepreneurs. Both liberation leaders and terrorists have created tight-knit communities of like-minded people.
Polarization is sometimes an excellent strategy. If the goal is to move people to a more radical position, and to entrench their positions, it's often a good idea to try to ensure that like-minded types are speaking mostly with, or least listening mostly to, one another. (Some student organizations are products of polarization entrepreneurs.) Authoritarian governments try to plan polarization, and sometimes they succeed; Orwell's 1984 and Animal Farm can be seen as case studies.
But sometimes -- and I think this is more interesting -- polarization arises spontaneously, through entirely voluntary choices, without the slightest kind of planning. Consider, for example, people's reading patterns, which suggest a kind of self-sorting into liberal and conservative networks. (See http://www.orgnet.com/divided.html) Or consider the blogosphere itself, where empirical evidence is starting to show a similar kind of spontaneous sorting and (in all likelihood) polarization.
Is this a problem? On the one hand, spontaneous polarization can lead people to more extreme positions, not because those positions are right, but simply because of limited information exchange and social dynamics. On the other hand, spontaneous polarization can increase what Heather Gerken has called "second-order diversity," in the form of a larger overall stock of social arguments and perspectives. Both of these effects are real, and in the abstract, it's hard to know which is more important.