If you’ve dabbled in online auctions or sales you will no doubt have wondered just how reliable are the various reputation ratings sellers report to convince you of their integrity. Interesting news this week from researchers at CMU who report that fraudsters in online auctions and selling sites can be reliably identified by the pattern of behaviour they reveal (http://tinyurl.com/yyr32m). The key seems to be identifying link patterns between group members, with fraudsters tending to display a clear pattern of communicating more with members in another group than the group of legitimate buyers and sellers on the site. This ‘bipartite core’ which can be clearly seen when the various transactions are plotted as a graph seems to be a strong indicator of the perpetrator working with accomplices to maintain a clean record on the original site while engaging in fraud with partners. The precise details are not yet revealed but most interesting is the emergence of this type of behavioral index from a massive data set using data mining techniques. What other indices of human behavior are out there to be uncovered is intruiging to consider. With Ebay reporting over 200 million registered users, the problem of small N for behavioral research might just start to appear quaintly old-fashioned.