Woj, you're a social scientist. How representative are self-selected, opinionated posters?
[WS:] We do not really know and it is evident that we do not know. With most "scientific" surveys, we do not know it either, but we fool ourselves that we do.
What we do know in scientific surveys is the distribution of the sample and population in terms of known characteristics, such as sex, age, residence etc. We do not know, however, the distribution in terms of unknown characteristics, say, political views, which we hope to assess (if we knew that, we would not be doing the survey!). We can only guess that such views are in a fixed relationship to the known characteristics, and on that basis claim representativeness, but such claim is only as good as our assumption. In fact, we do not really know. What we do know, however, is that those assumptions can be wrong, and the samples that we think is representative can be in fact biased.
In short, by browsing opinions posted to major media outlets, we can get an impression of the balance of public opinion without quantifying possible error (a serious limitation if you want to know, say, how votes will be cast.) However, since the major media outlets tend to attract broad audiences, there is a reason to believe that various views are represented. In scientific surveys we get an impression of what public thinks about the options given to them by researchers (a big limitation vis a vis open ended expressions as far as opinions on complex issues are concerned) and are able to quantify possible error, if , that is a big if, our assumptions about the relationship between known characteristics (such as socio-demographics, occupational status, residence etc) and the opinions we want to know hold.
In both cases, we are really guessing. Therefore, my counsel is to use both to develop a comprehensive view of reality. It would be foolish to dismiss one and believe only the other.
Wojtek