Isn't that something that Mendeley is trying to do? I would be interested in seeing something like this and helping when and where needed.
I've used Mendeley (and use papers.app when I'm on a mac) and it has helped me more with organizing papers than in filtering them. What you see as an academic in the morning is a 100 papers staring you in the face and you fall back on pattern recognition at points if you don't have more than a few minutes. Whose names and institutions do you recognize? Which titles seem most similar to your line of research? But you can miss a lot of interesting things and not view things tangentially related, but possibly useful as often. Journal clubs attempt to mediate this by having a group of people each pick an article and bring it to the table but crowdsourcing this and having cross lab "bringing to the table" would be very useful.