A novel perception-based DEA method to evaluate alternatives in uncertain online environments

Abstract

Consider a decision maker (DM) who must rank a set of alternatives and select one of them when searching online using a recommender engine such as Amazon or TripAdvisor. These websites provide numerical and linguistic reviews of the available alternatives offered by groups of unknown raters. The evaluations assigned by the raters to the characteristics of the different alternatives may or may not coincide with the evaluations that would be assigned by the DM if he were to actually observe the alternative. Hence, the value assigned by the DM to a characteristic must account for the uncertainty regarding the distribution of its realizations, the frictions inherent to the evaluations of the raters and the subjective quality of the perception determining his own evaluation. We formalize the incentives of the DM to select an alternative using a value function that incorporates these sources of uncertainty within a multi-criteria decision making environment. In addition, we implement this perception-based evaluation scenario within a data envelopment analysis (DEA) framework in order to study numerically the effects that perception differentials have on the ranking and selection behavior of the DM.