Multiple criteria decision support: The state of research and future directions

Multiple criteria decision making (MCDM) has been a popular research area for more than two decades. The interdisciplinary nature of MCDM, the pervasiveness of multiple criteria in decision problems, as well as good communication networks (several professional societies, working groups, conferences, summer schools, and newsletters) among others, have led to explosive international growth during the 1980s. Over the years several approaches and underlying theory have been developed in several countries around the world (Belgium, Bulgaria, Canada, England, Finland, France, Japan, the Netherlands, Poland, Portugal, Soviet Union, U.S.A., West Germany, etc.) for solving decision problems having multiple criteria. The types of problems addressed and the specifics of the approaches vary, but the ultimate goal is in common: “to help a decision-maker (DM) to find the ‘most preferred’ solution for his/her decision problem”. Conceptually, when the criteria and the decision alternatives are assumed to be known, the most preferred solution of a rationally behaving DM in set of decision^alternatives can be defined as a nondominated solution preferred by the DM to all other solutions at the moment of final choice. However, the definition is not very operational, because it is not realistic in general to assume that any MCDM-method could enable the DM to compare all possible solutions to the final solution at the moment of final choice. Therefore, MCDM-methods are always based on more or less restrictive implicit or explicit assumptions con~rning the DM’s preference structure and behavior. Any nondominated solution can be regarded as a rational choice, if there is no information about the DM’s preferences. In the 1970’s, research focused on the theoretical foundations of multiple objective mathematical programming and on the development of procedures and algorithms for solving multiple objective mathematical programming problems-especially linear and discrete problems. Mathematical problems dominated the field, and many ideas were adopted from the theory of mathematical programming. To characterize the Pareto optimal surface by enumerating nondominated extreme point solutions in multiple objective linear programming was a typical research problem. The use of an (implicitly or explicitly known) stable value function of certain form to model the DM’s preferences was a commonly used approach. There are still many mathematically challenging and important problems left. These include multiple objective integer, nonlinear, and stochastic optimization problems which further require mathematically oriented research. However, in the 198Os, the emphasis has clearly shifted towards multiple criteria decision support (MCDS). ‘I his shift means that research is focusing on the DM’s actual behavior (called behavioral realism), instead of solving a well-structured model under hypothetical and restrictive assumptions concerning the DM’s preference structure and behavior. The recently developed interactive multiple criteria decision systems (MCDS) which try to meet