The difficulty is that the "trolling" folks like Coates want to eliminate is often simply a difference of opinion which is acceptable in one community but not i

The difficulty is that the "trolling" folks like Coates want to eliminate is often simply a difference of opinion which is acceptable in one community but not in another. One of the few times I read Coates' blog was discussing a black/white gap in dating site response rates. He immediately banned people pointing out statistics showing that black women are disproportionately fat. The HN discussion on the same topic was heavily upvoted and rationally concluded concluded that explanation was probably incorrect. Building a system which will automatically detect that, e.g., statistics about race and obesity are trolling on Coates' site but useful contributions on HN is a difficult matter. Collaborative filtering is the best way to create a bubble, but that generally requires a centralized communication medium like Facebook, Twitter or Reddit.