And Prescriptive Purposes -some Comments*

The first problem encountered in discussions about the three closely linked purposes of decision research is a lack of agreement about the meanings of the three terms descriptive, normative, and prescriptive. The goal of these comments is to illustrate my understanding of the terms and suggest areas for future prescriptive research. Normative decision research identifies logically compelling properties with which decision behavior should conform. For a simple example, it can be forcefully argued that a person's preferences among alternative actions should be transitive. ~' Sets of compelling properties can be combined to identify various normative preference theories, such as expected utility theory as derived from axioms by yon Neumann and Morgenstern [6]. On the other hand, descriptive decision research identifies the way people actually make choices. Often, this research has explicitly compared people's choices with normative choices or models. Thus, continuing the example of transitivity, people have been shown to sometimes violate the transitivity principle (MacCrimmon [4,5]). There is a gap between the descriptive observation that people are sometimes intransitive and the normative principle that people ought to be transitive. Research with a prescriptive purpose is designed to bridge this gap by developing and testing methods for aiding people in conforming with desired normative principles. This can also be called decision engineering, since it requires a series of model building and testing phases. In his dissertation, Macerimmon found that verbally pointing out to subjects their intransitive orderings led people to readjust their orderings and become transitive. Transivity is a good introductory example since it allows a fairly clear-cut division among the three purposes for decision research. In many actual research projects, all three aspects may be present, though in different strengths. 1 profess to