1 March 2013, Computational Social Choice Seminar, Olivier Cailloux
I will briefly describe some typical goals of the multiple criteria decision aiding (MCDA) research field related to preference modeling, and as an example I will describe how this can be applied to the multicriteria portfolio selection problem.
MCDA is interested in situations where multiple points of view (criteria) may be used to rate objects (alternatives). We want to come up with an aggregated output, for example, a quality label for each alternative. When facing conflicting criteria, this requires a subjective appreciation of how the criteria interact.
Preference modeling is the activity which aims at building a model of the relation between the scores of evaluated alternatives on the criteria and the aggregated output. This models the preferences of a given individual (decision maker). The usual approach assumes that this relation is well-defined in the decision maker's mind, whereas MCDA, on the contrary, assumes that the decision maker can only talk about some imprecise constraints on the set of such possible relations. It is even possible that the constraints are in conflict. MCDA aims at coming up with methods to help individuals solve these conflicts and satisfy as much as possible their preferences.
As an example of what this different assumption permits, I will present an MCDA method to solve a multicriteria portfolio selection problem. Imagine a decision maker must establish a procedure to select a subset of persons according to their performance on a set of criteria (think about entrance examinations). The decision maker may like both the ideas of affirmative action (making sure the procedure benefits some usually underrepresented group) and of the non-discrimination principle (treat everybody equally). We will see how the proposed MCDA method can be used to help a decision maker ponder these two principles.