[andreareina]

This is neat. I think that {left, right, neither} is too broad a categorization, and suffers from the usual problem of self-reporting: most people are bad at judging themselves, and this is before you factor in the desire to paint themselves in a better light. An open affiliation could also make some people reluctant to participate (particularly with point (3) in your top-level post, seeing the distribution of posts upvoted).

Instead of having users explicitly choose from {left, right, neither} I would suggest not asking at all, and using analysis to find groups that vote the same way. You don't even need any labeling, just scores for how much the article agrees with the user's previous voting pattern and how broad of a group the article appeals to.

Now that I'm done insisting that the bike shed be painted blue I hope this takes off -- echo chambers do no-one good.

[mrbodoia]

Haha don't worry, I appreciate your input!

I agree that some kind of cluster analysis could be an interesting alternative to the explicit categorization that the current site uses. I thought that self-selection would probably be the best way to start, since clustering would be difficult without a large existing user base, but if the site takes off I'm definitely going to look into more advanced techniques for identifying partisan affiliation.

[weq]

Hey ive had a side-project in the works for a while now aimed in a very similiar manner to what you are trying todo here. I wouldnt mind throwing some ideas your way, as im not sure mine would get off the ground. Can u hit me up on weq at mooh org and we could chat a little?

[mrbodoia]

Sure, I'd love to hear your ideas!