Modeling Wikipedia admin elections using multidimensional behavioral social networks

Abstract

Wikipedia admins are editors entrusted with special privileges and duties, responsible for the community management of Wikipedia. They are elected using a special procedure defined by the Wikipedia community, called Request for Adminship (RfA). Because of the growing amount of management work (quality control, coordination, maintenance) on the Wikipedia, the importance of admins is growing. At the same time, there exists evidence that the admin community is growing more slowly than expected. We present an analysis of the RfA procedure in the Polish-language Wikipedia, since the procedure’s introduction in 2005. With the goal of discovering good candidates for new admins that could be accepted by the community, we model the admin elections using multidimensional behavioral social networks derived from the Wikipedia edit history. We find that we can classify the votes in the RfA procedures using this model with an accuracy level that should be sufficient to recommend candidates. We also propose and verify interpretations of the dimensions of the social network. We find that one of the dimensions, based on discussion on Wikipedia talk pages, can be validly interpreted as acquaintance among editors, and discuss the relevance of this dimension to the admin elections.