In the October 21 issue of Science, a News Focus article discussed using AMT as a source of experimental participants. The article included some exciting examples of research that was conducted on AMT that would have been difficult to conduct elsewhere, like a study that recently examined whether competence cues within politicians can be detected across cultures (download it here).
It also suggests something that many AMT experimenters may already know or suspect, but which has not been studied systematically yet: Different payments may attract different workers in terms of intrinsic and extrinsic motivations. Previous investigations found no evidence that low payments negatively affect data quality. This article suggests that, high payments might be likely to attract the wrong workers, with bad consequences, especially on tasks sensitive to intrinsic motivation. Not only may worker quality depend on the interaction between task and payment level, but it also may not be linear. These are additional considerations to be accounted for when considering payment level and should perhaps be systematically examined.
Access here the full text of the Science article.