Tools Supporting Choice

  • Andrzej P. Wierzbicki
  • Hirotaka Nakayama
Part of the The International Institute for Applied Systems Analysis book series (MMTA, volume 9)


There are two basic approaches to multi-objective decision support. One of these assumes that the main objective of a decision support system (DSS) is to support learning by the decision maker (DSS user, modeler); this was the motivation for presenting most of the techniques and tools in this book. A different and older approach assumes that the main objective of a DSS is to support choice.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. 1.
    Because many interpretations of classical approaches to decision analysis exist, we are aware that there might be some objections to classifying them as methods of “automating choice”. However, if we build a full model of a utility or value function of a human decision maker, we might just as well replace her/him by an automaton. In some cases, this is advantageous: automata do not tire and are always consistent.Google Scholar
  2. 2.
    This is a classical topic in decision analysis and much has already been written about it.Google Scholar
  3. 3.
    We should note that the concept of utility is reserved in decision theory for expected utility based on probabilistic representation of uncertainty. In cases without such representation of uncertainty it is preferable to speak of a value function.Google Scholar
  4. 4.
    Theoretically at least. Research in psychology and various paradoxes — such as the Allais paradox — show that one of the basic axioms of expected utility theory, the axiom of independence, does not necessarily describe the observed behavior of real decision makers. Thus, various alternative models of nonlinear dependence of expected utility on probabilities have been proposed (see e.g., French, 1986).Google Scholar
  5. 5.
    AHP is usually applied to problems with a discrete, finite number of decision options with directly specified outcomes or attributes. In principle, however, it could be applied to more complicated models, where the objectives yi are defined as (possibly nonlinear) functions of decisions x. Google Scholar
  6. 7.
    Also taking into account nondifferentiable and constrained cases.Google Scholar
  7. 8.
    The sign of this specific value as compared to the example considered in Chapter 6 is changed to make it easier to interpret positive quantities.Google Scholar
  8. 9.
    This is a very strong assumption; however, we use an arbitrary model to illustrate the diverse problems of investigating various model properties, and do not consider the issue of model validity.Google Scholar
  9. 10.
    A disadvantage of DIDAS-N is that it requires model formulation and modification in a traditional (not “user-friendly”) way.Google Scholar

Copyright information

© Springer Science+Business Media Dordrecht 2000

Authors and Affiliations

  • Andrzej P. Wierzbicki
  • Hirotaka Nakayama

There are no affiliations available

Personalised recommendations