An Approach for Modelling Preferences of Multiple Decision Makers

Modern decision making problems are discrete and multicriteria by nature, and involve several decision makers (DMs). One of the key questions in this type of problems is how the preferences of the DMs can be modelled. Usually the DMs are not sure of their preferences or will not tell them to the analyst, because they are not able to express their preferences directly. In these type of situations the decision support system should allow modelling of ignorance. Stochastic Multicriteria Acceptability Analysis (SMAA) is a family of methods to aid DMs in discrete decision aiding problems. Dempster-Shafer theory of evidence (DST) allows modelling of ignorance by using belief functions. In this paper we show how the preferences of multiple DMs or other stakeholders can be modelled and combined using DST, and how this information can then be encoded as interval constraints for sets of weights in SMAA. Our contribution has a wide range of practical applications. We present a real-world case study of human resource management in the public sector. It consists of evaluating the quality of education in the University of Turku, Finland.