Towards an Experimental Methodology of Virtual World Research

This paper presents the features that make virtual worlds particularly well suited for use as experimental environments, particularly, shards, dataveillance, the ability to manipulate the environment, and controlled random samples. It also attempts to identify and provide concrete examples of how researchers can avoid two key points of contention when using virtual worlds as experimental environments. The first point is that virtual worlds are not internally valid, and second is that data from virtual worlds cannot be generalized to draw conclusions about individuals in the real world. Finally, it introduces Greenland, an experimental environment that was developed from June 2008 to February 2009, and examines some early data from the project in hopes of translating it to some useful insights for others and for further development of the project itself.