The cover article of Science of July 27 is about the research potential of virtual worlds. It is written by William Sims (!!) Bainbridge, who is a program officer at the National Science Foundation. The argument is that nowadays so many people make use of virtual worlds, especially for entertainment, that it might be a platform to study social, behavioral and economic processes and derive high quality data for large populations. Although this sounds appealing, I have reservations in using this for the study of institutions and many other social phenomena since the population of participants is not a random sample, but a self-selected group of people who like to entertain themselves. Hence the motivations might be quite different if one likes to study how people share common resources.
Nevertheless, we should look at these developments seriously as it might provide a unique opportunity for the longer term to derive high quality data for large groups. I think this will require somewhat different virtual communities, and goals of the games beyond killing other avatars. At CSID we develop and perform experiments in virtual 2D and 3D environments. Our participant group is a random sample of our data base of undergraduate students at ASU and participants derive monetary incentives (one typically earns about $15-$25 in one hour experiments). I expect we will see in the coming years developments that will combine experimental environments with virtual worlds, but it will take quite some exploration to find out what the best format will be to be useful for testing all kind of social science theories.