✦ LIBER ✦

What can the regression model of human judgment learn from multi-attribute decision theory?

✍ Scribed by John Fountain

Publisher: Springer US
Year: 1978
Tongue: English
Weight: 270 KB
Volume: 9
Category: Article
ISSN: 0040-5833
DOI: 10.1007/bf00138122

No coin nor oath required. For personal study only.

✦ Synopsis

Zeleny's recent conjecture that multi-attribute decision theory may help to overcome the inadequacies of the linear regression model is incorrect. Recognition of the information processing advantages inherent in multiple-attribute decision situations combined with a requirement of transitivity itself implies linear objective functions. This follows from some recent developments by a psychologist and an economist in the analysis of individual and collective decision processes, developments which do not take as their starting point the paradigm of choice offered in utility theory.

In a recent criticism of the linear regression model of human judgment Milan Zeleny (1976) raised the question as to whether other developments in decision theory-multi-attribute decision theory in particular (Cochrane and Zeleny 1973)-might help to resolve some of the inadequacies of the linear regression approach. This paper argues that this question must be answered negatively. Quite the contrary appears to be the case. Recognition of problems peculiar to decisions involving multi-attribute alternatives supports the contention that the linear regression approach is a good one, methodologically speaking-in spite of the inadequacies pointed out by Zeleny. Why should this be so? Put succinctly, the answer must be that the decision maker who wishes to take account of the two requirements of information processing simplicity and transitivity must necessarily act as if he evaluates alternatives according to a linear objective function. Theoretical and experimental work on this question (in another context) can be found in Tversky . This paper presents the basic conceptual ideas and some illustrative examples. The mathematical proof can be found in Tverslcy (1969) or Fountain (1976).

1. INFORMATION PROCESSING SIMPLICITY FOR MULTI-ATTRIBUTED ALTERNATIVES

Consider the problem of a student trying to evaluate alternative graduate schools. Let Z be the set of graduate schools considered and suppose that any graduate school z eZ can be characterized according to a number of criteria. What is the size of its library? How much will they grant me as a stipend?