Capacity Refinements and Their Application to Qualitative Decision Evaluation
Résumé
This paper deals with the lack of discrimination of aggregation operations in decision-evaluation methods, typically in multi-factorial evaluation, and in decision under uncertainty. When the importance of groups of criteria is modeled by a monotonic but non-additive set-function, strict monotonicity of evaluations with respect to Pareto-dominance is no longer ensured. One way out of this problem is to refine this set-function. Two refinement techniques are presented, extending known refinements of possibility and necessity measures, respectively based on so-called discrimax and leximax orderings. Capacities then become representable by means of belief functions, plausibility functions or both. In particular it yields a natural technique for refining a Sugeno integral by means of a Choquet integral.