That gives me an idea. Given enough data, might it be possible to create a system that automatically identify "sides" to a controversial discussion based solely on how people are voting, and adjusts the sorting algorithm appropriately to counteract groupthink?

For example, if a group of people who tend to vote the same way on certain things all downvote a particular comment, but a smaller group of people with a different voting history vote a different way on the same comment, the system could mark that comment as a minority opinion rather than simply a "bad comment" and give less weight to downvotes from users of the opposing position when sorting that comment.