This project was design build and test computational theories describing the factors and processes that influence the success of online communities and test these theories again empirical data.   Although some communities succeed, the majority of newly created ones fail to survive or to achieve their goals to involve member or produce valuable artifacts. Even within a successful community like Wikipedia or the peer-to-peer lending site Kiva, some subgroups are more successful than others. One reason for failure is the lack of evidence-based guidance for building and managing online communities, as well as a lack of techniques to predict the effects of design and management decisions before implementation.  The theories developed and tested in this project predict community success at multiple levels of analysis ?members? support, community maintenance, production, and key stakeholder benefits.

This project used agent-based modeling, a computer simulation technique that models the decisions individual community members make in joining, contributing to or leaving the community. It will be based on and integrate component theories from social psychology, economics, organizational behavior and communications to model these decisions. It also built game-theoretic models to predict and test how team size and complementarity of members? skill predict the extent to which the community produces public benefits. The bullet points below describe some of the results from this project:

• Success in Wikipedia. We designed and built an agent-based model of to predict success of Wikipedia at multiple -levels, including the likelihood that new editors will continue participate, the number of articles produced and the quality of articles. This model was based on our literature review of success factors in Wikipedia and empirical studies showing when it its lifecycle article quality grows most quickly and how automated deletion of articles drives valuable editors away. We also conducted experiments to show ways to recruit experts to contribute to Wikipedia and to reduce the under-representation of women among Wikipedia editors. Experts are most likely to contribute when they are asked to contribute to topics that most closely match their expertise, and women are more likely to contribute when they are asked to write about a gender-neutral topic.
• Complementarity and team size. Do differences among team members? skills and other production-relevant attributes and the size of a team influence how much members contribute to a group outcome? For example, people who contribute money to a accomplish a group goals are relatively interchangeable, while those who contribute unique skills are complementary. We developed a theoretical model that predicts that for low complementarity teams (i.e., members substitute for each other), members in large groups contribute more whereas for high complementarity groups, members in small groups contribute more. This prediction is confirmed in our lab experiment. Moreover, our empirical research among Kiva lending teams shows that Kiva teams that in which members occupy a diversity of social roles contribute more than to teams where the members occupy fewer distinct social roles.
• Contributing a part of a group or individually. People contribute to some goal when they work with a group of other people than when they work alone. For example, joining a Wikiproject, which is a group of editors who curate Wikipedia articles on a defined topic, write more on articles that belong to the project but not to articles not connected to the project. Similarly, people who are experimentally encouraged to join Kiva peer lending teams contribute more money than those who do not get an invitation.
• Designing leaderboards: We built theory and conducted experiments that show that leaderboards, which show members of a group how well people are doing compared to each other, increase the amount that people work and that a fine-grained leader board with complete ranking outperforms a coarser leaderboard with just two categories.
• Identifying the roles members play in online communities. Using natural language processing and statistical models, we identified conversational acts that members perform when participating in online cancer support groups such as seeking informational support, providing empathy and self-disclosing information about themselves. We also identified 11 coherent social roles that members occupy in these groups, such as support provider, newcomer welcomer, and story sharer. Occupying certain roles such as newcomer-welcomer and encourager was positively correlated with members? sustained participation. In Kiva lending teams, we identified the conversational acts members perform, such as welcoming newcomers, encouraging others to give, reminding teammates of deadlines, or competing against other teams. Kiva teams that in which members occupy a diversity of social roles contribute more than to teams where the members occupy fewer distinct social roles.

Last Modified: 06/02/2021
Modified by: Robert Kraut