From Lurkers to Workers: Predicting Voluntary Contribution and Community Welfare

Abstract

In an online community, users can interact with fellow community members by voluntarily contributing to existing discussion threads or by starting new threads. In practice, however, the vast majority of a community’s users (90%) remain inactive (lurk), simply observing contributions made by intermittent (9%) and heavy (1%) contributors. Our research examines increases and decreases of types of user engagement in online communities using Hidden Markov Models. These models characterize latent states of user engagement from trace user activity or lack of activity. The resulting framework then differentiates lurkers who can later become "workers'' (i.e., engaged with the community) from those who will not. Differentiating lurkers who can be engaged from those who cannot enables managers to anticipate and proactively direct their resources towards the users who are most likely to become or remain workers (i.e., heavy contributors), thereby promoting community welfare. Analysis of 533,714 posts from an online diabetes community shows that incorporating latent user engagement variables can significantly improve the accuracy of welfare prediction models and guide managerial interventions. Application of our framework to five additional communities of various contexts demonstrates its generalizability.

Keywords: Online communities, Welfare of online communities, Voluntary online work, Predictive modeling