Researchers from Carnegie Mellon University's Human-Computer Interaction Institute have developed a statistical model for identifying potential administrator candidates, and propose an "AdminFinderBot" to automatically identify users likely to pass the Requests for adminship (RfA) process.
In their contribution to the 2008 Conference on Human Factors in Computing Systems, held in Florence, Italy this April, social psychologist Robert E. Kraut and Ph.D. student Moira Burke (User:Grammarnerd) describe the problem of increasing administrative backlogs (not currently a significant problem) and the shrinking ratio of administrators to non-administrators. They also argue that "people are more likely to contribute to a collective good such as Wikipedia when they know that they are uniquely qualified for the task, or that the likelihood of success is good", and that "[d]espite protestations that admins are lowly janitors mopping up, in many ways election to admin is a promotion, distinguishing an elite core group from the larger mass of editors." Thus, they suggest their model could be useful for both identifying editors with strong potential and as a "self-evaluation tool" for admin hopefuls.
The model takes into account users' detailed edit histories, including per-namespace edit counts, edits to specific page types such as deletion discussions, policy pages, WikiProjects and the administrators' noticeboard, and the presence of keywords (e.g., "POV", "revert") in edit summaries. For 2006 and 2007, during which 42% of adminship requests were successful, the model predicts with 75% accuracy whether candidates were successful or not.
According to the model, diversity of editing—and in particular, edits to Wikipedia policy pages, WikiProjects, and article talk pages—are the strongest indicators of likely success at RfA. Every 1000 edits to articles increases the probability of success by 1.8%, while edits to Wikipedia policy pages or WikiProjects have about ten times that effect. Every 1000 article talk edits boosts a candidates chances by 6.3%, while excessive userpage, user talk, and deletion discussion edits actually decrease chances for success. Comparing changes in the weighting of the model's factors from the pre-2006 period to the 2006-2007 period, the authors conclude that "the community as a whole is beginning to prioritize policymaking and organization experience over simple article-level coordination" when it comes to selecting administrators.
Another contribution to the conference, by researchers at the University of Pittsburgh and the University of Pennsylvania, analyzes the roles of rules and policies on Wikipedia. The authors conclude that, despite the reputation of wikis as venues for "peer-based, nonhierarchical, non-bureaucratic, emergent, complex, and communal" work, the key feature of wikis is that they "allow for, and in fact facilitate, the creation of policies and procedures that serve a wide variety of functions."