This is kind of what I tried to do many months ago. It was also based on the premise of transparent authority, many-to-many content and user connections, catego

This is kind of what I tried to do many months ago. It was also based on the premise of transparent authority, many-to-many content and user connections, categorization on multiple dimensions, and such. I also approached it from a sociolinguistics angle. The problem with sociology and the like is that the metrics themselves are very broad and hard to quantify. The application's complexity spiraled out of my control. I still believe in the core concepts, but execution of this thing in particular is very tricky. Theoretically speaking it could be the end-all recommendation engine, and that is, from my own experience, almost as lofty as building a Turing Machine. Granted that you have user input, it probably only needs to be half as strong, but that is already very strong, and very difficult (at least for me). Now logic is one thing. One the same level of difficulty is how to make it user friendly; for this app, regardless of how hard the AI is, the UI is probably half the battle. I'm not sure what kind of app you have in mind specificially. dcurtis's mockup is real slick and to the point. My implementation was a lot more passive, but I suspect internally the structures share commonalities with your proposal. If you are interested we can discuss. I won't be applying with my implementation though, because this the kind of site where without 100k users, a VC won't even bother. And no matter how good it is, it is much harder to get traction today, hence higher uncertainty, higher burn.