Wikis and work groups: A social network approach to predicting community growth (Foote, Shaw, Hill)

Large peer production projects, such as Wikipedia and open source software, work surprisingly well in creating useful, large scale, and high quality artifacts. However, the vast majority of peer production projects fail to gain contributors or contributions. Because network scholars have focused on the rare successful projects, there is very little research on the factors that predict project growth in the first place. We approach this question by examining the network structures and participation dynamics of a diverse population of peer production communities as they are just starting.

Early stage peer production communities resemble two types of collaboration networks: work groups engaged in information sharing and voluntary attempts at collective action. The literatures on these topics provide very different predictions about what sorts of factors should be important in predicting community productivity and growth. In order to better understand the determinants of effective collaboration in peer production, we test predictions from both literatures in a large population of peer production projects. According to models of collective action, individuals make sequential decisions about how much to contribute to a collective good based on the current status of the good’s production. This literature predicts that projects which have a critical mass of dedicated contributors, and which are growing quickly, are likely to grow quickly in the future. In other words, higher tempo and larger scale participation are positively associated with building collective action.

Alternatively, the performance of work groups engaged in information sharing tasks depends on the structure of participation. The work group literature has found that dense, non-hierarchical interaction structures are associated with more productive group performance. Because both work groups and new peer production projects are small groups of individuals engaged in collaborative sensemaking to produce a shared artifact, we might expect that these same structures would also lead to more productivity from new peer production communities.

To test these claims, we create collaboration and communication networks for 2,555 wiki communities hosted on the website Wikia.com. At the point when each wiki had received 500 total edits we construct measures of early-stage network structure and participation dynamics. We find that early interaction structures have very little impact on eventual productivity (as measured by total edits) or community growth (as measured by total contributors), once we control for the size of the networks and how quickly the community is producing content. Rather, we find that the speed at which content is being produced, as well as the number of active editors predict both productivity and growth. These results provide evidence that early peer production projects look more like collective action than like work groups