Thursday, June 07, 2007

On Blog Commenting



While I was writing my response to Joe Klein's post about blogging I started thinking about something else, and that is the selection bias in blog comments, including those that are submitted at the Time's own blogs.

A "selection bias" is my borrowing from the literature on surveys in statistics. One problem with sampling a group of people and then generalizing from their opinions to some wider group is that unless we are very careful the people in the sample might not look like the people in the wider group. For instance, if you ask about people's opinions in some suburban mall you are going to have only people who visit malls, only people who are well enough to go out and very few rural people. So your findings are not necessarily going to reflect what those omitted groups think.

Similar problems apply to all those Internet polls. People self-select to participate in them, and those who find the time worth wasting are those who feel most strongly on the issue that the poll asks about. In general the more extreme opinions will be overrepresented in such polls, and that is why they always come with that warning about not being scientific. Which means that they actually are pointless things to have.

Now, comments on blogs are not intended to be a random sample of opinions but too often they are treated that way. Yet for me to comment on someone else's blog requires that I'm interested enough in the topic of the post, that I have the time to spend writing a comment, that I think I know enough about the question and often my emotions also need to be engaged.

What does this mean in terms of, say, the anger of the blogs? The people who are the angriest are most likely to comment, I suspect, and mostly it will be comments which are critical of the post, because just saying "hear, hear!" doesn't seem quite as necessary to do as expressing an opposing viewpoint. After all, the post itself expressed what you agree with. More controversial topics and topics which have to do with something we all feel we know about are also going to get more comments than, say, a post on research findings concerning the efficacy of abstinence education.

Based on my estimates from my own readership figures, the number of blog readers who actually comment is a lot less than one in a hundred. Yet the blogosphere is often characterized by the types of comments that appear, and this is where the fact that the commenters are not a random sample should be stressed.

In short, the people who read, say, progressive blogs, are not necessarily like the people who comment on those blogs and the two should not be treated as the same.