Predicting Edit Locations on Wikipedia using Revision History

There has been increasing interest in the machine learning community in automatic task design. In a collaborative problem-solving setting, how can we best break up and assign tasks so as to optimize output? Huang et al., for example, considered the problem of effectively assigning image-labeling tasks to Amazon Mechanical Turkers [1]. In the realm of Wikipedia prediction, Cosley et al. created a successful system for recommending articles a user might be interested in editing [2]. Nicknamed SuggestBot, the system makes use of co-editing patterns and articles the user has previously edited to predict new articles of interest. This system represents a success in automatic task design: deploying SuggestBot increased a user’s edit rate an average of fourfold.