Doherty, P., Driankov, D., and Tsoukias, A. (1992). Partiality, ┼ara-consistency and Preference Modeling : Preliminary Version. Technical Report LiTH-IDA-R-92-18, Department of Computer and Information Science, Link÷ping University, Sweden. (bibtex),
Abstract: Existing non-monotonic formalisms are generally designed to capture a single criterion for determining preference among competing models. As pointed out by Jon Doyle, one must aggregate individual preference orders into a global order to obtain unified non-monotonic formalisms. As long as one considers preference modeling in terms of classical decision theory, any attempt at aggregating different criteria will violate certain commonsense postulates about preference orderings. Fortunately, recent advances in multiple-criteria decision theory permit the use of preference orderings not limited by the classical constraints of complete comparability and transitivity. The advantage is that one is able to model decision scenarios where information is incomplete and inconsistent. In addition, non-conventional preference modeling may provide a potential solution to the aggregation problem discussed by Doyle. One obstacle towards applying non-conventional preference modeling has been the lack of a sufficient axiomatization for non-conventional preference structures. The main contribution of this paper is to propose the use of a multi-valued para-consistent logic, WL4, as a means of providing a robust semantic theory for non-conventional preference modeling.
CS Dept TR Overview