An ongoing thread of research questions whether there is some political risk to hanging out and talking to people who are pretty much like ourselves. The theory is that we're better off with a network of folks who disagree with our points of view. This will expose people to other viewpoints. But the reality is we tend to have interpersonal networks that look a lot like ourselves.
Some folks tested whether it matters in terms of political knowledge -- is your conversational network made up of people who agree with you, or people who disagree.
The result? It "suggests that individuals who reside in diverse networks benefit little from frequent political talk, whereas those surrounded by like-minded discussion partners appear to acquire significantly more knowledge as a function of increasing discussion." The article appeared in a 2008 issue of Communication Research.
At first I found this counterintuitive, and there are some caveats. The authors use issue knowledge and it has a lousy Cronbach's Alpha of .47. That's troubling, so much so I'm a little surprised the research got through the journal's vetting process. Their background knowledge index performs significantly better.
I don't want to go all PhDweeb on this one, but the authors have some neat suggestions on why they found what they did. Maybe negative or non-supporting conversations confuse people to the degree that knowledge gets all messed up. Perhaps affect gets in the way of learning. This is odd since there are some suggestions that advertising, even negative adds, improve learning. As scholars often do, almost as boilerplate, there is a call for more research to tease out what's going on here.