Phased multicriteria preference finding

Abstract

This paper considers the implications of a tendency of multicriteria decision-makers to use screening, ordering and choosing phases to find a preference as they reduce the set of their candidate alternatives. This corresponds with increasing levels of effort and willingness to use sophisticated cognitive processes. It suggests that appropriate corresponding types of measurement are those using attributes, utilities and relative scores of alternatives, for which standard methods are SMART, MAUT and AHP. Theory on the decision processes of the mind is incorporated, which shows how to structure criteria trees and that relatively measured weights and scores should be synthesised using a power function. Tests of revised utility and relative MCDM methods and of software incorporating these ideas are reviewed. They facilitate interactive refinement of scores at all points of the criteria tree and easy reduction in the number of alternatives. Two alternative screening phases are considered, one based on scoring all the attributes together, the other scoring the attributes in clusters in the structured criteria tree. Empirical tests confirm the value of the three phase approach, but leave slightly open the question about which type of first phase is preferable.