Mining user behaviors in large social networks

GIVEN large social networks that have recorded millions of users and items (eg, messages on Facebook/Twitter, articles on Shashdot/Digg), billions of item adoption behaviors and millions of user connection behaviors, what is the underlying behavioral mechanism of the social network user? Can we accurately predict the most probable items that the user will adopt? Can we recommend to them what they really want? How can we alleviate