Premium
Experimental consideration of preference in decision making under certainty
Author(s) -
Corner James L.,
Buchanan John T.
Publication year - 1995
Publication title -
journal of multi‐criteria decision analysis
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.462
H-Index - 47
eISSN - 1099-1360
pISSN - 1057-9214
DOI - 10.1002/mcda.4020040204
Subject(s) - judgement , certainty , ranking (information retrieval) , preference , multiple criteria decision analysis , rank (graph theory) , sample (material) , value (mathematics) , computer science , decision maker , decision analysis , operations research , artificial intelligence , management science , mathematics , machine learning , mathematical economics , statistics , engineering , epistemology , chemistry , geometry , chromatography , combinatorics , philosophy
This paper describes an experiment in decision making under certainty with multiple, conflicting objectives and continuous decision variables. Two techniques for analysing such problems are considered: one taken from the paradigm of multicriteria decision making (MCDM), a non‐directed approach called the NAIVE technique, and one from the paradigm of multiattribute decision analysis (D/A), the SMART technique. While the two techniques seek and are throught to arrive at the same end—a solution which is in some sense optimal for the decision maker (DM)—the former approach implicitly incorporates DM preferences while the latter approach considers preferences explicitly. The setting is a laboratory study using a sample of university students on a three‐criteria problem which is designed to study the extent to which value functions implied/assessed by the techniques are consistent with DMs' holistic ranking of alternatives. Results show that (1) the two techniques of interest show significantly better rank order correlation with holistic judgement compared with other techniques, (2) DMs prefer the non‐directed MCDM approach and (3) subjects break down into two groups: those that use assessable value functions when ranking and those that do not. This implies that for small‐dimensioned problems DMs may first need to be classified as to the assessability of their value functions before a solution method is chosen.