INTRODUCTION

In the previous chapter we discussed the strategic choices that the analyst has in conducting an analysis of a decision. While these strategic approaches vary considerably and on many dimensions, there a number of skills (or tools) that the analyst needs to complete any one strategy, and these skills are amazingly common across all strategies.

First in Section 7.2 we will talk about structuring, by which we mean both the definition of the problem and the identification of the elements of the model to be used in analyzing the alternatives. The choice of an analytic strategy is included in problem-structuring. Also included in this section is some practical advice on how to get the analysis started and a philosophical, yet pragmatic, view of modeling.

The third section of this chapter addresses the use of iterative analysis, the sequential application of structuring and quantification. Four types of iterative analysis are discussed, followed by a discussion of how to use iterative analysis in deciding how much analysis to do.

Probability elicitation occupies Section 7.4. Research findings on good and bad ways to assess probabilities are summarized before an encoding process is described. Finally we comment on the discretization of continuous probability distributions for decision tree analysis.

THE DIFFICULTIES OF DECISION-MAKING

In some ways we can describe decision-making as the most common human activity. Almost everything we do involves decisions. Whether we are driving a car, preparing a menu for a meal, working out what to say next in a piece of writing, or handling a difficult social situation, to give just a few examples, we have to make choices and decide to adopt one course of action rather than another. Since we are involved in so much decision-making, we might expect that the human species has evolved to be good at it. Viewed in one way this is indeed the case, in that we have developed many innate mechanisms for effective intuitive decision-making in our best interests. As babies, our unconscious decision to cry when hungry, or in pain, surely is the wisest way of improving our lot! As mature adults, our instinct to dodge a stone thrown at us is an intuitive decision that maximizes our welfare. These are at the unconscious level. At the conscious level, however, we have also developed ways of making decisions without much thought. Habits of driving, acting according to conventions of social behavior, following norms of hygiene or diet, all lead to good decision-making.

After defining what we mean by rationality in decision-making in Chapter 2, we give an account of decision theory in Chapter 3. Chapter 4 consists of a review of psychological research in decision-making, and we discuss organizational decision-making in Chapter 5.

INTRODUCTION

As we have already remarked, decision-making is a subject which attracts the attention of scholars in numbers of different disciplines. In practical fields where decision-making is a frequent and crucial activity, such as management, medicine, town planning or engineering, much has been written on the subject. The theoretical literature is even larger, however. Mathematicians have studied the implications of the axioms of decision theory; statisticians have been concerned with how to decide in the face of uncertainty; economists have studied how human decision-making causes, and is caused by, economic activity; philosophers have grappled with what it means to make rational decisions. It is to psychologists that we turn, however, for a consideration of some of the central questions of decision-making. These are:

(1) What causes us to make a decision? What are the determinants of decision-making?

and, more usefully:

(2) How can we make better decisions?

Psychologists have given considerable attention to these questions since the 1950s; the purpose of this chapter will be to describe some of this work and its implication for the practice of decision analysis.

Good reviews of the literature are available elsewhere (see Edwards 1961, Slovic et al. 1977, Einhorn and Hogarth 1981, Pitz and Sachs 1984 or Slovic et al. 1986).

Having addressed the abstract and theoretical issues related to rational individual and organizational decision-making, we now turn to the practice of decision analysis. Chapter 6 addresses the strategic (and stylistic) approaches to implementing decision theory for real problems. In Chapter 7 we present the analytic skills (or tactics) needed to practice decision analysis. Then these skills are illustrated with real applications in Chapter 8. The final chapter discusses the validation of analyses, the ethics involved in practicing decision analysis, and research areas in the field that need attention in the future.

In this closing chapter we will appraise in three ways the decision synthesis that we have described in the other chapters. First we will discuss how we can know whether to trust the recommendations of a decision analysis — should the synthesized decision be taken, or ought we to be cautious about the limitations of our analysis? This is the study of how a decision synthesis may be validated, and it will be discussed in Section 9.1. This will involve an account of how sensitivity analysis may be used to improve the validity of an analysis and, in a rather different argument, discussion showing that, if decision synthesis is constructive, validity is less of a problem. Secondly, in Section 9.2, we shall discuss what needs to be done to make decision synthesis a yet more effective procedure for enhancing decision-making. We shall review the research areas that, at the time of writing, seem to be the most important, and speculate on the likely changes that we might see in the practice of decision analysis in the next few decades. Finally, we conclude the book with a discussion of the ethics of doing decision synthesis.

THE VALIDATION OF DECISION ANALYSIS

When first introduced to the ideas of decision theory, many people find the structure and consistency provided by the theory very attractive, but then find great difficulty in applying the ideas to practical problems.

Decision Synthesis makes up and lives up to its title. The authors have done an exemplary job of piecing together the diverse strands of the emerging, multidisciplinary field of behaviorally oriented decision theory in a way that serves their stated goal – improving decision-making by synthesizing techniques and perspectives with a sensitivity to the peculiarities of individual decision problems and decision-makers. The text draws upon diverse literatures whose creators generally know about one anothers' existence, but never have quite found the time to work out the interrelationships between their respective approaches. Although there are too many of these pairwise relationships for many to be fleshed out in any great detail, seminal ideas can be found at many junctures. As such, it is a book that gives some of the future of decision-making, as well as well-chosen pieces of its past and present.

In creating this synthesis, Watson and Buede have adopted a decidedly disorderly approach to the often tidy world of decision theory. They present rudiments of the important normative axiomatizations of formal prescriptive approaches. However, they also caution the reader against taking it all too literally. Following much current thinking, they thoughtfully discuss the limits to imputing descriptive validity to normative theories – that is, assuming that people make decisions the way they are supposed to.

In the previous chapter we argued that, to assist people in thinking through complex decision problems, a quantitative framework is needed to encourage consistency, the essence of rationality. Decision theory provides such a framework, and in this chapter we give an account of this theory. We start, in Section 3.1, with value function theory, which is concerned with conflicting objectives. In Section 3.2 we discuss uncertainty, and how the calculus of probability may be used to describe it. Section 3.3 contains a description of utility theory, which is concerned with decision-making in the face of uncertainty, and in Section 3.4 we show how uncertainty and conflicting objectives may be handled at the same time. Finally, in Section 3.5, we discuss other approaches to creating a calculus for decision-making.

CONFLICTING OBJECTIVES

We all make decisions all the time. What clothes shall I wear this morning? What shall I have for lunch? Shall I watch television or read a book? Most of us have no difficulty in deciding matters like these, but on occasion we find ourselves spending a lot of mental effort in making up our minds. Choosing a career and then a particular job, or where to live, often falls into this category.

This is a book about decision-making – about how people do make decisions, and about how they may improve their decision-making. To say this makes some assumptions, and, before we go on to show the reader how decision-making will be tackled in this book, as we will do in the rest of this introductory chapter, we will examine these assumptions.

First, can decision-making be studied? One of the authors, when discussing his research work with a senior (and traditionally minded) colleague, was asked what subject he studied. On replying that he studied decision-making, the response was “Is that a subject?” Many readers will not need our prompting to answer that question affirmatively, but it is as well to spell out the nature of our study. It is principally concerned with the determinants of decision – that is to say, what it is that makes us take one course of action rather than another. In some cultural and intellectual traditions there is an easy answer to this question. Extreme forms of Calvinism, for example, itself a particularly severe kind of Protestantism, deny that free will exists; nobody may make decisions, therefore, since everything is fore-ordained by God. Then, as Howard (1980) points out, in an excellent article on the philosophy of analyzing decisions, many Eastern philosophies will not see the study of decision-making as a fruitful endeavor, since in these philosophies one accepts the unfolding of life as it presents itself.

Each of the chapters in this volume concerns some aspect of economists' use of controlled experiments. Since the mid-1970s this kind of work has been transformed from a seldom encountered curiosity to a small but well-established and growing part of the economic literature. This transformation has been rapid. For example, when I began my own experimental work about a dozen years ago, it was most convenient to publish the results in journals of psychology and business. Today it is no longer unusual for controlled experiments to be reported in any of the major American economics journals. Experimental work has become well enough represented in the literature so that, in 1985, the Journal of Economic Literature established a separate bibliographic category, “Experimental Economic Methods.”

However, as might be expected of any newly developing field of scientific endeavor, there are at least as many points of view about the role of experiments in economics as there are economists who conduct them. One of the reasons for this is that “economics” encompasses quite a diversity of activities and methodologies, and controlled experimentation appears to have the potential to play at least a supporting role, and in some cases a far larger part, in many of these.

Introduction

The term “parallelism” refers to a vague notion about how observations of simple laboratory phenomena can help one understand and predict the behavior of a complicated and changing world. Of what use are experimental results to someone who is interested in something vastly larger and more complicated, perhaps fundamentally different than anything that can be studied in a laboratory setting? Questions such as this and the related notion of parallelism have probably existed from the earliest development of scientific experimental methodology, and although I found the term in a paper by Vernon Smith (1980) the notion itself pervades all branches of science and engineering.

The purpose of this chapter is to isolate some examples of how the issue of parallelism has been approached in economics. The chapter outlines several strategies that have been employed in attempts to use experimental research in actual policy decision making. The topic to be explored is how issues have been posed in these policy-related studies so that experimental methods could be applied. The discussion is limited to 10 instances in which I have been involved personally at some level.

Many different opinions exist about experimental methodology and the relationship between laboratory work, field studies, and policy decisions.

Introduction

Games in characteristic function form were introduced by von Neumann and Morgenstern (1944). Laboratory experiments on such games have led to descriptive theories of coalition bargaining. No theory proposed up to now is completely satisfactory in light of the data. However, the evidence clearly suggests that equity considerations have a strong influence on observed payoff divisions. The purpose of this chapter is to elucidate this phenomenon.

The formal structure of equity considerations can be expressed by an “equity principle,” which is explained in Section 3.2. This principle is well known in the social psychology literature (Homans, 1961; Adams, 1963; Leventhal and Michaels, 1969; Walster, Walster, and Berscheid, 1978; Harris, 1976; Mikula, 1980). The terminology used here is based on a paper published elsewhere (Selten, 1978).

To some extent the influence of equity considerations on the behavior of subjects in coalition experiments may be due to the subjects' desire to conform to social norms. However, a different explanation of the phenomenon seems to be more adequate for most of the experimental results.

In a unanimity game where the players can either all agree on the division of a fixed sum of money or else end up with zero payoffs for everyone, the inherent symmetry of the situation points to equal shares for all players.