On Subjective Trust Management

Abstract

Trust and reputation management is gaining nowadays more attention then ever as online commodity exchange and other open virtual societies became a widespread reality. Most widely used computational models use reputation metrics as global property assigned to each party. More sophisticated models try to use reputation as subjective property. While introducing subjective reputation there arise a need to model preferences of agents. In this paper we propose to use weighted ordered weighted average (WOWA) operator to support the decision maker in assessing available evidence about other’s party behavior. The WOWA aggregation is defined by two weighting vectors: the preferential weights assigned to the ordered quantities and the importance weights assigned to several attributes. It allows one to express both the preference regarding sources of information by the corresponding importance weights and the compensation between attribute values aggregated by the preferential weights.