This is very true, but I feel that the group think isn't so much mediated by style as it is by what people are actually saying; this is a harder problem to solv

This is very true, but I feel that the group think isn't so much mediated by style as it is by what people are actually saying; this is a harder problem to solve. :-( Respectful discussion prevents people from flying off the handle and getting distracted from the main topic. Me-too'ers and yes-men are too agreeable, but still on topic. I wonder if it's possible to analyze a paragraph for agreeableness and automatically reduce its visibility if it's too extreme on either end. That might force people to post a more balanced view (assuming that the people gaming the system are down-voted by human moderators). This actually seems like something a machine could reasonably do. Agreeableness is vague enough that it wouldn't need domain knowledge of the conversation, and there are probably enough key words and phrases to train it. I wonder what effect that would have?