Basically, the study works like this. The avatar first makes a purposely unreasonable request out of a stranger. Then, when it is refused, they ask another, more moderate request (this tactic -- typically used in experiments such as this -- is done to try and increase response rates).
In one of the most striking findings, the effect of the DITF technique was significantly reduced when the requesting avatar was dark-toned. The white avatars in the DITF experiment received about a 20 percent increase in compliance with the moderate request; the increase for the dark-toned avatars was 8 percent.
The nice thing about studies like this is they can abstract away a lot of subtle, personal tics which can confound normal person-to-person studies. The behavior of virtual avatars can be standardized far more than can an individual person (or group of persons).
A more extensive summary of the study can be found here. The citation is: Paul W. Eastwick and Wendi L. Gardner, Is it a game? Evidence for social influence in the virtual world, 4 Soc. Influence 18 (2009).