Encoding monotonic multiset preferences using CI-nets

Abstract

CP-nets and their variants constitute one of the main AI approaches for specifying and reasoning about preferences. CI-nets, in particular, are a CP-inspired formalism for representing ordinal preferences over sets of goods, which are typically monotonic. Considering also that goods often come in multisets rather than sets, a natural question is whether CI-nets can be used more or less directly to encode preferences over multisets. We here provide some initial ideas about this by first presenting a straight-forward generalisation of CI-nets to multisets with bounded multiplicities, which we show can be e ciently reduced to CI-nets. Second, we sketch a proposal for a further generalisation which allows for encoding preferences over multisets with unbounded multiplicities, yet characterise reasoning in this framework in terms of the first. We finally show a potential use of our generalisation of CI-nets for personalization in a recent system for evidence aggregation.