Modelling of preferences and multicriteria decision aiding

The objective of this research project is to develop and study the different steps of a decision aiding process that integrates the preferences of a decision maker. It includes theoretical aspects (development and axiomatic study of new preference models), methodological aspects (development of multicriteria methods and elicitation techniques) and practical aspects (implementation of these methods on real cases). The project brings together concerns from Operations Research and Artificial Intelligence. Its 4 axes are as follows:

1. Problem structuring: we are interested in the general form of a decision aiding problem (very often of a multicriteria nature, with criteria that are often conflicting), as well as in the notions that are strongly associated with it, for example the way in which the set of alternatives or the family of criteria is defined and constructed. We reflect on basic decision support concepts such as compensation and/or incomparability.

2. Preference structures:* Once the decision problem is well-structured, it becomes important to study the nature of the preferences encountered in the development of the decision model. Thus, this axis focusses on the modelling of sophisticated preferences (with thresholds, biorders, families of relations), preferences on the Cartesian product (joint measurement), the axiomatisation of preference aggregation functions and the elicitation of preferences.

3. Validation: This important aspect of solving a decision aiding problem consists of validating the methodology used in the decision process and the recommendations provided to the decision maker. The questions we are particularly interested in here are: which decision models are the most executable? how to define and find the traceability of the multi-criteria decision process? how to use argumentation theory to justify the recommendations?

4. Application: Study of ranking procedures and design of indicators. For example: bibliometrics, social networks, hospital ranking.

*: point 2 is a theme that is in strong interaction with the project Games and social choice: axiomatic and algorithmic aspects

Keywords: preference modelling, joint measurement, multicriteria methods, ELECTRE, UTA, elicitation, evaluation, indicators, argumentation, validation