NORMATIVE-PRESCRIPTIVE VERSUS BEHAVIORAL-DESCRIPTIVE THEMES IN JUDGMENT AND DECISION MAKING

Perhaps more than other areas of the human sciences, JDM research includes elements of both description and prescrip­tion, of trying to discover what people actually do when they form judgments and make decisions and of advising them on how they might do these things better. The advice-giving theme can be traced to mathematicians of the eighteenth cen­tury French court who offered advice on such matters as the fair price for gambles (Bernstein, 1996; Stigler, 1986). The roots of the descriptive theme are more widely scattered but were well established by the time of two landmark review pa­pers (Edwards, 1954, 1961) that substantially launched be­havioral interest in decision making.

The two themes seem to be built into the subject matter. If one starts, for example, with an interest in how a doctor makes a particular difficult diagnosis (e.g., Einhorn, 1974), one would probably investigate the types of diagnostic in­formation that the doctor collects, the way she puts it to­gether into an overall judgment, her ability to reproduce the same judgment on repeated cases, and so on. But it would be hard not to ask the evaluative questions: How well is she doing? Are her diagnoses correct? How well could anyone, or a computer, do in making this diagnosis from this information? How might she be helped to do it better?

Conversely, a decision analyst might be able to show that given specified preferences and probability estimates manager would be well advised to make a given set of in vestments. This still leaves open the manager's ability t₀ state appropriate preferences and to assess required proba­bilities—and to generate enough faith in the entire analysis to be prepared to take action based on it. Thus, serious de­scriptive work on decisions often reaches important norma­tive questions, while intendedly prescriptive studies rise or fall on the realism with which they represent the psychology of the decision maker.

This interplay of descriptive and prescriptive issues is a central source of interest to many JDM researchers. How­ever, it has also led to what many see as an undue interest in decision errors. A major research program of the 1970s and 1980s, associated with Kahneman and Tversky (see the later section on heuristics and biases), assumes that observed decision behavior is generated by a reasonably small number of cognitive rules of thumb or heuristics, mental shortcuts that generally produce reasonable (and quick) results. These heuristics were demonstrated by showing that people generate systematic "errors" in spe­cific, carefully constructed situations. The errors were de­fined as a deviation between what a subject did and the conclusions derived from some optimal rule—for example, a subject's probability estimate when given some informa­tion and the estimate that would be generated by Bayes's theorem in the same situation. This investigation of errors took on something of a life of its own (Edwards & von Winderfeldt, 1986; Jungermann, 1983), ignoring the facts that (a) the errors existed only if the optimal rule was, in fact, appropriate and accepted, and (b) there was little effort to assess the generality of the errors.

None of this is to suggest that humans are immune to de­cision error. Most of us, drawing on scientific evidence and personal experience alike, are happy to accept any help that is offered in our important life decisions. It is not clear, how­ever, how common serious decision errors actually are. How might one assess an overall decisional batting average for the typical human, other than citing casual evidence suggesting that it is close to neither 0 nor 1,000? Without an agreement on what constitutes decision error and an overall estimate of its frequency, one cannot assess how serious the biases caused by heuristic use might be. We argue only that when presented with a normative recommendation, it is always wise to ask if its assumptions are descriptively accurate and that when presented with a descriptive analysis of some deci­sion maker, it is always interesting to ask how well he or she is doing.

Date: 2016-03-03; view: 666