Researchers from Stanford and Cornell Universities have developed an algorithm that only needs to observe five to ten online posts to predict whether a member of an online community needs banning