I wrote a paper arguing that synthetic worlds had already performed interesting quasi-experiments; it was published in the prestigious journal Games and Culture. The point was that if game developers were unconsciously generating interesting experiments at the social level, just think what could be done if researchers became developers. The main experiment in the paper involved market location dynamics. The theory that I claimed EverQuest confirmed has just been tested again, in WoW, and happily it may have just been confirmed again. We would call this an out of sample confirmation. When it happens, it's generally viewed as strong confirmation that the method and the theory really do sit together nicely. Synthetic worlds really are good tools for social experimentation. I'll explain, below the fold.
The theory is that market location is an n-player asymmetric coordination game. When EQ started, markets were in EC tunnel, North Freeport, or Greater Faydark. Then there were server splits, and the pattern of splits showed that coordination effects, rather than mere history or economics, were the dominant predictors of where the markets went. That's the argument of the Games and Culture paper.
WoW used to have Auction Houses only in one city per side. Focusing on Alliance (Horde sucks and is evil, as everyone knows, so I will ignore it), the city was Ironforge. Then WoW opened up Stormwind and Darnassus for auctions as well. The prediction would be that if coordination effects are strong, one city would be used. If Ironforge remained as the main market, that would be inconclusive - maybe it is just history. But if it remained one city, but not Ironforge, that would confirm the power of coordination.
I'm told by David Bricker of the Indiana University Presidential Adminstration that Stormwind has become the main market on the Alliance side. I'm not in Wow myself at the moment. If this is true, and generalized across servers and the Good/Noble vs Evil/Suckage divide, this would generally confirm the theory. And it would be more evdience that large games are good for experiments.