Readings in Multiple Criteria Decision Aid

Preferences play a fundamental role both at the individual and at the collective levels. For this reason, preference modelling is an important step in decision-aid, economy, sociology, psychology, operations research,…

In Multi Criteria Decision Making one is generally concerned with decisions under certainty, i. e. decisions for which the “state” is assumed to be known with certainty. Multi Criteria Decision Making under risk or uncertainty would imply the super-imposition of the problem structures of classical MCDM and that of single criteria decision making under risk, i. e., for instance, the combinations of goal programming with stochastic programming. This would, obviously, become very involved mathematically! In this paper we are not concerned with uncertainties (probabilities) of the Kolmogroroff type but rather with uncertainties as they are considered in the theory of fuzzy sets, possibility theory and the like. It will be shown that for this type of uncertainty (vagueness) which is assumed to be more relevant for MCDM, models and methods exist, which are also adequate for MCDM and which are computationally still feasable.

The Analytic Hierarchy Process (Saaty, 1977 and 1980) is a multiple criterion evaluation methodology that is both descriptive and prescriptive. The Analytic Hierarchy Process (AHP) is, in many ways, similar to Multi Attribute Utility Theory. However, unlike MAUT, AHP does not prescribe that judgments be perfectly consistent, nor does it prescribe when or when not to allow for rank reversals. AHP allows the decision makers to decide how much inconsistency is reasonable, if any, and whether nor not rank reversal (a reflection of relative rather absolute worth) should be permitted.

V·I·S·A is a computer program for multiple criteria decision aid, based on a simple weighted multi-attribute value function, incorporating a hierarchical structure of criteria and visual interactive sensitivity analysis. The use of a model of this kind as an aid to multi-attribute decision making is nothing new. Long before MCDM became an established field of study in the 1970’s use of models of this kind been reported in the literature, for example, Churchman and Ackoff, 1954. Many popular approaches to decision aiding adopt the same framework, for example, Edwards’ SMART (Edwards, 1982), the Analytic Hierarchy Process developed by Saaty (Saaty, 1980), Social Cost Benefit Analysis (Lichfield et al., 1975). The approach is best suited to the problem of choosing a preferred alternative from a set of well defined alternatives, or to indicate a preference ordering over such a set of alternatives. There are many other approaches suited to this type of problem, for example, the Electre methods and Promethee. However, few approaches have been developed to handle a situation in which the decision maker wants to take account of very many criteria; in such a situation those methods incorporating a hierarchical structure of criteria are most appropriate, namely the multi-attribute value function and the Analytic Hierarchy Process.

This paper discusses 32 topics that summarize interactive multiple objective programming, the area represented by STEM by Benayoun, de Montgolfier, Tergny and Larichev (1971), the Geoffrion-Dyer-Feinberg Procedure (1972), the Zionts-Wallenius procedure (1976 and 1983), Wierzbicki’s Reference Point Method (1977, 1982 and 1986), etc. The concensus is that a spectrum of interactive procedures is necessary because the procedure to use on a given application is usually problem and user dependent.

STRANGE as an original approach has been already described in the literature (Teghem et al., 1986, Teghem and Kunsch, 1988, Slowinski and Teghem, 1988). It is a method designed to treat multiobjective and stochastic linear programming (MOSLP) problem. Some industrial applications has been solved using this method (Teghem and Kunsch, 1985, Kunsch and Teghem, 1987).