more on eurekster
Peter Caputa has obviously been experimenting with eurekster (and talking about it in his eureksterblog!) longer than I, and he has a number of good observations. Responding to my post, he mentions that they're still trying to decide what degree of friend to use for re-ranking search results. I wonder if they've considered using decayed influence. I.e., my friends have a lot of influence, their friends have somewhat less, and so on out to my four- or five-hop friends who have very little influence (at least individually; collectively they might still guide me toward the meme of the moment).