Finding Optimal Policies for Online Communities with CoSiMo
Felix Schwagereit, Sergej Sizov, Steffen Staab · 2010
The rapidly increasing popularity of Web 2.0 online communities originates in the ease of collaborative content creation and its sharing. As a result, more community members actively participate in the community and its data growth rates are continuously increasing. This poses the challenge for community platform operators on efficient administration and moderation to ensure the quality of content and to prevent violations of laws (e.g. copyright, privacy, illegal content) and community rules. Involvement of employed administrators who read and approve every piece of user-generated content is clearly the safest way of quality assurance. Since this is a time consuming task it does not scale up with Web 2.0 dimensions. So administrative functions are delegated to members of the community, the moderators. The strategy for choosing trustworthy moderators in big anonymous communities is specified in policies based on user reputation that is measured in bonus points. The proper balancing between community self-management and administration is crucial for the quality, attractiveness, and scalability of the entire community. Therefore understanding the mutual influences between community actors, reputation systems and platform policies that employ user reputation is crucial to ensure the overall quality, user acceptance, and success of the entire online community. Our objective is to predict the behavior in an online community for different policies, which may result in different overall quality of the community content. For this purpose we present our community analysis framework CoSiMo (an acronym for Community Simulation and Modeling), which employs the model-based approach for predicting the impact of policies on community dynamics and health. Through systematic variation of simulated quality assurance mechanisms we show that our model plausibly captures the influence of policies to content quality and can be therefore exploited for optimization of real online communities.