PSYCHOMETRIKA--VOL. 54, NO. 2, 363--364 JUNE 1989 REVIEWS

REVIEWS D. von Winterfeldt and W. Edwards. (1986). Decision Analysis and Behavioral Research. Cambridge: Cambridge University Press. pp. 604. This book is informative, accessible, and unusual. It should be on the shelf of every scientist who is interested in how people make decisions (i.e., behavioral decision making) or how people help other people to make decisions (i.e., decision analysis). The book attempts an integration of the principles derived from the study of human judgment and decision making with the tools that have been developed to improve decision quality. It succeeds quite nicely. It is informative because the coverage is broad and, often, deep. The first of the 14 chapters provides an overview of decision analysis and basic issues. The second is devoted to the frequently neglected problem of decision structuring. Much of this chapter seems somewhat horticultural, dealing with the growth and shaping of a veritable orchard of trees fault trees, event trees, inference trees, and value trees. The third chapter extends the arboreal metaphor with a detailed and delightful discussion of the development of decision trees. The following three chapters deal with uncertainty. Chapter 4 introduces central concepts of probability. Independence, exchangeability, the elicitation of likelihood judgments, and the issue of calibration are included. This chapter is followed by one introducing Bayesian ideas, most of which are rather stale, and by another one, entitled "Inference", which is devoted almost exclusively to complex, multistage, cascaded inference structures. Chapters 7 through 10 deal with preference issues. The first of these contrasts value and utility measurement and presents a useful discussion of techniques for scaling preferences. The ideas are extended to multi-attribute value structures in the following chapter where different models are described and compared. Chapter 9 develops the measurement theory for both the preceding chapters. The attention paid here to the various independence assumptions is especially useful since these properties often constitute the testable implications of the models. The final chapter of this section reviews empirical research bearing on the descriptive adequacy of the unidimensional and multi-attribute models covered in the preceding chapters. Prospect theory is introduced here as an alternative to expected utility theory. The first of the final four chapters discusses sensitivity analysis. Some interesting insights are offered here about the nature of the "flat maximum" phenomenon, the tendency for a value or utility function to be relatively flat in the neighborhood of its maximum. This chapter also includes an interesting discussion of the robustness of linear regression models and unit weighting. Chapter 12 provides useful summaries of individual cases of decision analyses along with authors' accumulated wisdom on how to avoid the multitude of pitfalls that lurk in the "bushy messes" of decision analysis. Chapter 13 presents an original view of research on cognitive biases, illusions, and heuristics along with some thoughts on the relationships among the concepts of intuition, cognitive processes, and intellectual gadgetry (calculators, computers, and other mental aids). There is a lot more to think and argue about in this chapter than in the final 363

364

PSYCHOMETRIKA

one, which offers an idiosyncratic (and rather autobiographical) history of decision making ideas in psychology and thoughts on the future. The book's accessibility is a remarkable achievement that results from at least four features. First, the writing is uniformly clear, simple, and graceful. Second, the style is refreshingly informal and, often, personal. The decision tree that is grown and pruned in chapter 3 concerns a decision one of the authors had to make about suing an uninsured motorist for an automobile accident. This chapter also illustrates the third quality that makes the book so readable, and that is the lavish use of examples to illustrate conceptual points. (There is, for instance, an example of a set of preferences that satisfy the single but not the joint independence assumption of conjoint measurement. What a nice way to make the distinction concrete.) Finally, the book is packed with figures (over 125) and tables (over 65) that aid comprehension. What makes the book unusual is that it not only tries to weave concepts from psychological research into decision analysis, it also stresses the need to combine the science of decision analysis with the human sensitivity that is needed when consulting with a client. So while the core of the book focuses on the description and application of mathematical models of decision making, the book's secondary message is that the best applied science may be futile if the social, organizational, and/or political context of the application is ignored. The book deals with both the science and the art of decision analysis in other words. This virtue, however, may limit the suitability of the book as a text in courses in which the art of decision analysis is deemed superfluous. It is possible to find faults with the book to be sure. There are the inevitable, and inevitably annoying errors in equations and figures here and there, but they are infrequent and generally easy to spot. I was disappointed in the lack of coverage of decision analysis in situations with multiple decision makers (only one example is given in chapter 12 of an analysis in which the preferences of three interest groups are represented), but I was heartened to read in chapter 14 that this is a promising area for future development. I would have liked to have seen more than just cascaded inference theory in the treatment of complex inferences, and when the authors write that they know of no work on syllogistic reasoning since World War II, I conclude that their library needs updating. These shortcomings do not seriously diminish the book's major accomplishment, however, which is the thoughtful blending of psychological research into decision analysis. UNIVERSITY OF CALIFORNIA, SANTA BARBARA

David M. Messick