Man, I'd love to have that as a project! For reddit, I wrote a pretty simple program that checked reddit every few days an sorted through the connections and began to rank users: It ranked who was the most "successful" link submitters (in terms of amount of votes compared to number of items submitted) as well as the most "successful" commenters (in terms of comment-to upvote ration). For commenters, it also looked for trends with other commenters. I found that this was only really useful in smaller subreddits where individuals could actually emerge as being known by their reputation (this is ignoring "power users" who mostly become famous through promiscuity alone). It was rather interesting to see that users could mostly only really be "influential" in smaller subreddits/communities where individual merit could be noticed.
keep me in the loop I could use some better methods of data extraction.
Well, like you'd said: this largley depends on there being an API for hubski. Reddit has an awesome API that's simple to jump right into (it uses JSON, a pretty intuitive way of storing data, as well as a drop dead simple way of accessing it). Without it, I'd pretty much be forced to do page scraping, and that's just not a lot of fun (and it's made a bit tougher because most user data is actually retrieved via javascript in the page and is dynamic, not stored on a static page). I can't wait for the Hubski API to come out!