Have you ever thought about trying to visually map an argument or discussion? I find trying to follow arguments or discussions in threads really confusing and timeconsuming. I'd love to be able to look at a visual representation of a discussion. The way I'd imagine it would work is that highly rated or highly dispute comments would be taller. Differences of opinion would then cause a branch in the discussion. maybe red for points of dispute, green for points of agreement. It would be fascinating to watch one of these 'visual arguments' without actually having to read the discussion. A group of people who just agree with one another would just be a long straight green line. A group who disagree would create lots of red and then split into two split green branches.

I once was part of a site/community which took opinions or points or view, and mapped their relationships and had pie charts and such.  Unfortunately, it could not overcome its flaws, and it just degenerated for lack of moderation and inability to offer a proper medium for intelligent discussion.

Ultimately, it boils down to quality of content, and not so much the gimmickry.   It might be hard to automate your proposed system, as well.  Would human beings be the ones to build the data and connections?
 

This is a great idea. But I can't see how it would be possible -- maybe with another 10 years of natural language technology.

I can see this being very interesting to analyze. And the technology I'm sure could be implemented with a few rules in place. Here is a good post from Lois Kelley on language analysis on her blog: http://blog.foghound.com/192/

Anthony

Most discussion forums have some sort of threads and ratings for posts already so you can do some form of visual mapping already.

Yes, that's true. I thought you mean more automated sentiment analysis (I've been reading too much on this subject)

I might try and find a heated debate somewhere and visually map it

Try the Science vs Religion group

I think there are a lot of ways to show conversations. Especially with some of the Java based knowledge maps out there showing number of replies = size of topic you don't need to get so much into 'understanding' the content itself.  Also individuals are pretty consistent in their ability to be argumentative so organizing by respondent might yield some interesting results.

Another thing.. aren't dissenting posts usually longer than supportive one's? That might also be an easy way to group things. More and longer post replies might mean a more interesting conversation, especially ones where the longer posts are later in the thread.
 

I think that's a correct and pretty interesting observation about discussion analysis. Would be some interesting experiments worth running there. Though I think we might leave that more to the community once we release an API.

You might enjoy having a look at a new interest group on Freebase on the topic of argument modeling (a larger topic than argument visual mapping, but clearly related)