In this paper, we use activity traces to try to cluster users based on their behavior in an online genealogy wiki. We treat the wiki as a community of practice, and based on legitimate peripheral participation theories we expect that people will start in peripheral roles before moving to more central roles.
We use agent-based network simulations (RSiena) to model how and whether people change roles over time. We ran into a few difficulties in the project - first, it was difficult to identify meaningful roles using k-means clustering. There was a very peripheral, low-activity role and two central roles but we couldn’t identify any moderately central roles. We also had problems with the agent-based modeling. The networks we identified changed too quickly for RSiena to model it effectively.
This is a paper I’d like to get back to and try to publish sometime but I know it will require digging into an old and ugly codebase so I’ve avoided it. :)