> I think a new killer app would be a form of automated shill-detection for forums... the trouble would be reducing the false positives to zero. An interesting

> I think a new killer app would be a form of automated shill-detection for forums... the trouble would be reducing the false positives to zero. An interesting idea for AI! Although, the domain knowledge and contextual information required to classify a poster as say, "normal vs. astroturfer", might be too great, and vary from topic to topic. Also, there's the question of how positive / negative training examples would be obtained - who's going to admit to the ground-truth labels for the "astroturf" class, for example?