Dimensions of Reputation in Electronic Markets
Google Tech Talks
September 25, 2007
his is based on a pair of papers. The idea is that reputation systems in electronic markets capture complex and multidimensional information about seller ability. Buyers are also likely to value different dimensions differently. Simply reporting an average score seems narrow. The research identifies different dimensions by mining reputation text feedback, clustering it and scoring it using actual premiums in price sellers transact at. The first study estimates a series of econometric models to show how price variation can be explained better by this richer view of reputation (it also provides evidence that buyers actually care about different things). The second study builds a predictive model that uses the theory and techniques to predict future sales on Amazon.com.