Dealing with uncertainties in MCDA

Abstract

This chapter presents various approaches to incorporating formal modelling of risks and uncertainties into multi-criteria decision analysis, in a theoretically valid but also operationally meaningful manner. We consider both internal uncertainties (in the formulation and modelling of the decision problem), and external uncertainties arising from exogenous factors, but with greater attention paid to the latter. After a broad discussion on the meaning of uncertainty, we first review approaches to sensitivity analysis, which is particularly, although not exclusively, relevant to internal uncertainties. We discuss the role, but also some limitations, of representing uncertainties in formal probabilistic structures, linked also to concepts of expected (multi-attribute) utility theory. Such probability/utility approaches may be used in explicitly identifying a most preferred solution, or simply to eliminate certain courses of action when stochastically dominated (in various senses) by others. In some contexts it may be useful to view minimization of various risk measures as additional criteria in more standard MCDA models, and we comment on advantages and disadvantages of such approaches. Finally we discuss the integration of MCDA with scenario planning, in order to deal with deeper uncertainties (not easily if at all representable by probability models), particularly in a strategic planning context. The emphasis throughout is on the practice of MCDA rather than on esoteric theoretical results.