Thierry Marchant

We characterize the orderings of pairs of sets induced by several distances: Hamming, Jaccard, Sørensen-Dice and Overlap. We also characterize these distances.

ELECTRE TRI-nB is a method designed to sort alternatives evaluated on several attributes into ordered categories. It is an extension of ELECTRE TRI-B, using several limiting profiles, instead of just one, to delimit each category. ELECTRE TRI-nB comes in two flavours: pseudo-conjunctive and pseudo-disjunctive. In a previous paper we have characterized the ordered partitions that can be obtained with ELECTRE TRI-nB, pseudo-conjunctive, using a simple axiom called linearity. The present paper is dedicated to the axiomatic analysis of ELECTRE TRI-nB, pseudo-disjunctive. It also provides some combinatorial results.

The paper provides an axiomatic characterization of a family of rank dependent weighted average utility criteria applicable to decisions under ignorance or objective ambiguity. A decision under ignorance is described by the finite set of its final consequences while a decision under objective ambiguity is described by a finite set of probability distributions over a set of final consequences. The criteria characterized are those that assign to every element in a set a weight that depends upon the rank of this element if it was available for sure (or non-ambiguously) and that compare sets on the basis of their weighted utility for some utility function. A specific subfamily of these criteria that requires the weights to be proportional to each other is also characterized.

The paper provides an axiomatic characterization of a family of rank dependent weighted average utility criteria applicable to decisions under ignorance or objective ambiguity. A decision under ignorance is described by the finite set of its final consequences while a decision under objective ambiguity is described by a finite set of probability distributions over a set of final consequences. The criteria characterized are those that assign to every element in a set a weight that depends upon the rank of this element if it was available for sure (or non-ambiguously) and that compare sets on the basis of their weighted utility for some utility function. A specific subfamily of these criteria that requires the weights to be proportional to each other is also characterized.