Sat 2nd October 2010, 3:13pm
QUOTE(Peter Damian @ Sat 2nd October 2010, 10:02am)
Are we reading the same paper? The goal of the paper, as stated in the abstract, was to build a model that will predict who will succeed at RfA, and to use 'policy capture' to determine how well the policy matches the actual promotion. As I read the paper, there are 'similarities and differences' - some aspects of policy are captured, others aren't (I may be wrong, I didn't read it that carefully).
The silliest aspect of the paper was the idea of using the model to build an 'adminfinderbot'.
That's because it comes from CMU. Surely you're familiar with the sort of work that goes on there; much of what they do focuses on robotics, automation, artificial intelligence, and statistical decision engines. This is right up that alley.
I did read the paper, and at several points they noted that factors that Wikipedia claims increase one's chance of passing are either neutral or negative predictors of passing. In addition, the authors are clearly skeptical of Wikipedia's claims about its own processes; they're not suffering from Reagle's unwillingness to see past the pretty wrapper. They even used the term "level up" to refer to seeking adminship, and they consistently refer to voters as "voters", with a footnote commenting on how Wikipedians refuse to call them that, but they're going to use that term anyway for "simplicity".