An agent-based model to understand tradeoffs in online community design

In this paper, we built an agent-based model to integrate and synthesize social science theories to inform online community design. As an example, we use the model to examine how various styles of moderating online conversations affect community performance. We compared three styles of moderation - no moderation, in which all members are exposed to all messages; community-level moderation, in which off-topic messages are deleted for the group; and personalized moderation, in which different messages are shown to different individuals to match their interests. We examined the effects of these moderation techniques in communities with various topical breadths and message volumes. Virtual experimental results suggest that despite the widespread use of community-level moderation, personalized moderation was more effective in increasing member commitment and contribution, especially in communities with a broad set of member interests and high message volume. Compared with no moderation, community-level moderation increased members' likelihood of reading but not posting messages.