I don't think he's claiming the variation in content has decreased, but warns that dependence on collective rating systems could decrease the variance of an individual's effective sample size.

I also don't think the author claims the web has introduced any social problems, just that it isn't solving them yet. Are these even problems with technical solutions?

He seems to be ignoring the "long tail" argument that the internet, with its vast reaches, enables many people in weird niches to find one another.