The MAUT-Machine: An Adaptive Recommender System

Abstract

In this paper we introduce an adaptive recommender system that supports the user in finding interesting entries in an electronic product catalog. The Analytic Hierarchy Process (AHP) lays the foundation for the user preferences. The user interests are specified in terms of desired properties of the ideal product, expressed in the form of constraints (or criteria) on the products basic attributes. Logically related criteria can be combined to form a criteria tree and weights are used to specify the relative importance of these criteria (their contribution impact on the overall rating). An extension of the Multi-Attribute Utility Theory (MAUT) that allows complex and powerful combinations of criteria is used to compute the degree of interest (or utility) of the products regarding the user preferences. Once the user has defined her preferences she can quickly identify the most promising products in the treemaps display of the catalogue, colour-coding schemes indicating the different degrees of interest and the user having the possibility to sort the catalogue by several of its products basic attributes. The user profile and the catalogue are simultaneously visible and each modification of her preferences is immediately mirrored in the treemaps display that always presents the quality of the catalogue entries w.r.t. the current user preferences. The user can also select a given criteria node in her preferences and see the performance of the catalogue entries for that criteria. This way, rational decision making with multi-criteria objectives is significantly alleviated.