Structure of Decision: The Cognitive Maps of Political Elites

by Robert Axelrod (Editor)

Overview

This book outlines a new approach to the analysis of decision making based on "cognitive maps." A cognitive map is a graphic representation intended to capture the structure of a decision maker's stated beliefs about a particular problem. Following introductory chapters that develop the theory and techniques of cognitive mapping, a set of five empirical studies applies these new techniques to five policy areas.

Originally published in 1976.

The Princeton Legacy Library uses the latest print-on-demand technology to again make available previously out-of-print books from the distinguished backlist of Princeton University Press. These editions preserve the original texts of these important books while presenting them in durable paperback and hardcover editions. The goal of the Princeton Legacy Library is to vastly increase access to the rich scholarly heritage found in the thousands of books published by Princeton University Press since its founding in 1905.

Product Details

ISBN-13:	9780691644165
Publisher:	Princeton University Press
Publication date:	04/19/2016
Series:	Princeton Legacy Library , #1707
Pages:	422
Product dimensions:	6.10(w) x 9.30(h) x 1.10(d)

Read an Excerpt

Structure of Decision

The Cognitive Maps of Political Elites

---

By Robert Axelrod

PRINCETON UNIVERSITY PRESS

Copyright © 1976 Princeton University Press
All rights reserved.
ISBN: 978-0-691-10050-0

---

CHAPTER 1

The Cognitive Mapping Approach to Decision Making

This chapter introduces the cognitive mapping approach to decision making. It describes what cognitive maps are and how they can be empirically derived. Finally, as an indication of how the cognitive mapping approach can help in the understanding, and ultimately in the improvement, of decision making, a chapter-by-chapter overview of the entire volume is presented.

IMPROVING DECISION MAKING

This is a book about how people make decisions and how they can make better decisions. It focuses on an analysis of the effects of policy alternatives upon valued goals. Our approach is to study the cognitive maps used by actual decision makers and policy experts, to see in what ways their performance can be improved. The theme of the book is that although people have to simplify their image of the complexities of the environment in order to cope with it, the number and types of simplifications need not be as great as they usually are. The tools we have used to establish this can also be used to help future decision makers.

This may sound like an ambitious program, and it is. But not impossible. After all, decision makers are typically either experts in a more or less specialized policy domain (such as transportation or foreign policy), or they are "specialists" in aggregating diverse interests (such as Congressmen and the President). Decision makers, however, are not necessarily aware of the specific shortcuts they use in thinking about policy choices. And even if they are sometimes self-conscious about their own cognitive styles, they may not be aware of the limitations these styles impose, and how these limitations could be ameliorated.

Here is a loose analogy. A Roman supply officer employing Roman numerals would not have given much thought to the structure of the system of calculation he was using. He would have thought about the quantity of supplies he was purchasing, their quality, their price, their delivery time, and so on. But he would probably not have considered the fact that his system of arithmetic calculation could be substantially improved by the use of Arabic numerals. If someone had tried to show him that the use of Arabic numerals would make his task considerably easier, his first response might have been that he could already perform all of the calculations he needed to do in his job, and that the new system looked strange and probably was not worth the trouble it took to learn, assuming it worked at all. In a way, the official would be right, because he could already do all that was expected of him ("his best"), and any new system would take some effort to master.

But in retrospect, we can say that the Arabic numerals are a definite improvement over Roman numerals, because they make many tasks, such as long division, substantially easier. Moreover, a person operating with Arabic numerals would have a much better feel for what he was doing. In Arabic numerals, for example, the number with the longer representation is the larger number, but this is not necessarily true in Roman numbers (as in 901 vs 8 and CMI vs VIII). Finally, a place number system such as the Arabic system is superior because it makes it easier to develop more advanced techniques, such as decimals. What lies behind the superiority of the Arabic system over the Roman system is that it better reflects the underlying characteristics of the tens-based number system used by both systems. Indeed, the reasonably skilled user of Arabic numerals might look like a prodigy in a world of Roman-numeral users.

Of course, the evaluation of policy choices is much more complex than the solution of an arithmetic problem, but this is all the more reason why there may be room for improvement. If we had a better idea of the methods people use to analyze and evaluate complex policy choices, then we would be in a better position to design a formalized system to help them do their analysis. Such a formalized system, reflecting the underlying characteristics of the way people actually do evaluate complex choices, could be expected to provide three significant benefits:

1. it would make certain types of evaluation easier;

2. it would expand the range of complexity that a reasonably skilled person could handle; and

3. it would promote the development of new ways of thinking that would be logical extensions of the existing methods.

THE ANALYSIS OF CAUSAL ASSERTIONS

This book proposes such a system, based on the notion of causation. The notion of causation is vital to the process of evaluating alternatives. Regardless of philosophical difficulties involved in the meaning of causation, people do evaluate complex policy alternatives in terms of the consequences a particular choice would cause, and ultimately of what the sum of all these effects would be. Indeed, such causal analysis is built into our language, and it would be very difficult for us to think completely in other terms, even if we tried.

The basic elements of the proposed system are quite simple. The concepts a person uses are represented as points, and the causal links between these concepts are represented as arrows between these points. This gives a pictorial representation of the causal assertions of a person as a graph of points and arrows. This kind of representation of assertions as a graph will be called a cognitive map. The policy alternatives, all of the various causes and effects, the goals, and the ultimate utility of the decision maker can all be thought of as concept variables, and represented as points in the cognitive map. The real power of this approach appears when a cognitive map is pictured in graph form; it is then relatively easy to see how each of the concepts and causal relationships relate to each other, and to see the overall structure of the whole set of portrayed assertions.

For added rigor, a mathematical system is presented for the analysis of these cognitive maps. With this system it is possible to derive formal inferences, which can be applied in many ways, such as seeing whether people actually make choices that are consistent with their whole collection of stated causal assertions. Such inferences can also be used to offer advice based on a given cognitive map of the implications of specific choices.

THE DERIVATION OF COGNITIVE MAPS

The analysis of a given cognitive map is one thing, but its actual derivation is another, and often far more difficult, task. For the study of decision making, the methods used to derive a cognitive map should, as far as possible, satisfy four quite demanding requirements.

First, the methods should be unobtrusive. To guard against the possibility that policy makers could not or would not reconstruct the relevant set of causal assertions after the fact, the cognitive map would have to be derivable from whatever materials are left behind in the normal course of a decision-making process.

Second, the derivation should not require advance specification of the concepts a particular decision maker may use in his cognitive map. Ideally, these concepts, as well as the causal links between them, would come from the data and not from any a priori assumptions of the researchers.

Third, the derived cognitive map should be closely tied to an evaluation theory of decision making, so that it can be used to advise and even criticize the decision maker. This means that for a given policy issue, the options, the goals, the ultimate utility, and the relevant intervening concepts should all be included in the cognitive map of a decision maker. Moreover, the evaluative theory of decision making that is employed should be sufficiently sophisticated as not to beg the important questions involved in reaching a complex policy decision.

Fourth, the method for deriving the cognitive map should be valid, which is to say that the map should be an accurate representation of the assertions (and relationships among them) used by the decision maker. One important aspect of the validity of such a method is its reliability, that is, the capacity of the method to yield the same results when used by different researchers under the same conditions.

Social scientists have already developed several techniques for the study of people's assertions, so one might well consider these older methods before undertaking the development of new procedures. Survey research techniques, for example, have achieved very good validity by any reasonable social science standards, and they can be tied to an evaluative theory of decision making, as they have in the study of voting behavior. But survey techniques are hardly unobtrusive, and they have only limited capacity to derive concepts that are unique to each respondent. Content analysis, another measurement approach, has achieved very high reliability and good validity in certain contexts. Moreover, content analysis techniques can be used unobtrusively, since they are based on documents, and they have considerable flexibility for adopting the categories used by different people. Unfortunately, even the most fully developed form of content analysis is still essentially a counting procedure with limited usefulness for analyzing the structure of the relationships between the concepts. Finally, the open-ended probing interview is an extremely flexible data-gathering method, but it is hardly unobtrusive and has limitations in reliability.

While each of these established methods has its advantages for the derivation of a cognitive map, a new method was necessary to take advantage of the opportunities that arise when suitable documentary material is available. Deriving a cognitive map from existing documents has the advantage of being both unobtrusive and fully able to employ the concepts used by the decision maker himself. Moreover, with the use of the analytic techniques of cognitive maps, the information can be tied to an evaluative theory of decision making. The first three empirical studies of this volume use a coding method for the derivation of the cognitive maps which they analyze. The validity of the documentary coding method is not yet fully established, although the empirical studies that use it offer different kinds of evidence in support of its validity. Reliability, however, is fully established. After more than three years of work, the coding rules for deriving a cognitive map from a document have reached a state of precision such that intercoder reliability is fully compatible with the accepted standards of good quantitative work in the social sciences.

An alternative method for the derivation of a cognitive map is used in the remaining two empirical studies. This method employs a questionnaire sent to a panel of judges who are in a position to make informed estimates of causal links. The questionnaire method is necessarily at least one step removed from the decision-making process itself. Its chief advantages are that it allows for the aggregation of individual opinions, and that it may be based on a much wider range of information than research can select for documentary analysis.

These two methods are used to generate comparable cognitive maps, and therefore the same kind of analysis can be used on a cognitive map derived from either method. It turns out that there are interesting differences in the properties of the maps derived by these two methods, and these differences reveal important facts about the cognitive capacities of decision makers.

A third potential method for deriving a cognitive map is the use of an open-ended probing interview. This method could also be used to generate comparable cognitive maps, and it has the advantage of allowing the researcher to interact actively with the source of his data. Work is already in progress by two of the contributors (Bonham and Shapiro) in the use of open-ended interviews to generate cognitive maps.

COGNITIVE MAPPING AND OTHER SYSTEMS

The cognitive mapping approach to decision making uses elements from at least four fields.

1. Psycho-logic. The idea that a mathematical system can be specially designed to deal with a person's cognitive processes may be traced to Abelson and Rosenberg (1958), who call their system "psycho-logic." Their system also uses points and arrows, but they give a different interpretation to them. Using Lambert's terminology (1966, p. 126), the points in psycho-logic are "'thing-like' concepts that for a given subject [person] are relevant to a given object." In cognitive maps, on the other hand, the points are not things, but variables that can take on different values. Thus, for example, in a cognitive map a person is never a point, although some of a person's perceived properties (such as his wealth and utility) may be points. This treatment of points as variables rather than as things makes a cognitive map an algebraic rather than a logical system.

The arrows are also treated differently. In psycho-logic the positive and negative arrows "may be viewed as equivalent, respectively, to the English expressions 'is usefully or desirably associated with' and 'is adversely or undesirably associated with.'" (Lambert, 1966, p. 126). In a cognitive map, the arrows are not representations of attitudinal association, but rather are representations of causal assertions about how one concept variable affects another. Although the interpretations of the two systems are different, from a strictly mathematical point of view, a cognitive map can be regarded as a generalization of psycho-logic. In particular, the ideas of ambivalence and balance will be taken from psycho-logic and extended for use in cognitive mapping,

2. Causal Inference. The idea that points can be regarded as variables, and that the arrows can be regarded as causal connections between the points, comes from the statistical literature of causal inference, developed by Simon (1957), Blalock (1964), and others. The task of causal inference is to provide a statistical technique to estimate the parameters appropriate to describe a given body of data, subject to a certain set of assumptions about the error terms and the structure of the causal links. That is a long way from the purpose of cognitive mapping, which is to represent what people actually say about causal relationships. While some of the basic ideas of casual inference have been quite useful in this volume, the very complex calculations that are typically involved in causal inference obviously bear little resemblance to what a person is able to do in his head.

3. Graph Theory. Both psycho-logic and causal inference employ some of the mathematical ideas of graph theory, as does cognitive mapping. Graph theory provides concepts, such as paths, cycles, and components, that are helpful in the analysis of complex structures of interconnections. Cartwright and Harary (1965) have shown how Heider's (1946) original idea of cognitive balance among three objects can be generalized to any number of objects, and Harary, Norman, and Cartwright (1965) have shown how psycho-logic can be represented in graph theoretic terms. Cognitive mapping uses graph theory, but generalizes it by allowing the points as well as the arrows to take on different values. The resulting mathematical system is similar to the network system developed by Maruyama (1963) to analyze mutual causal relationships in the environment.

4. Evaluative Assertion Analysis. The ideal that structural relationships between pairs of concepts can be systematically and reliably coded from a document was taken from evaluative assertion analysis of Osgood, Saporta, and Nunnally (1956).

5. Decision Theory. From decision theory have been taken the ideas of choice and utility. (See, for example, Luce and Raiffa, 1957). One of the intended contributions of cognitive mapping is to help analyze the decision-making process in terms of the structure of the relationships that a person asserts connect his choices with the expected outcomes of these choices.

The overall research strategy of this volume is to base what is being measured on what is being asserted rather than what is being thought by a person. One basic tactic is to infer things about decision making and cognition from a systematic study of these assertions, and especially from a study of the structural relationships among these assertions.

This research tactic can usefully be contrasted with another recent approach to the study of cognitive processes, namely, the artificial intelligence approach. In the cognitive mapping approach to cognition, the researcher starts with linguistic behavior and makes inferences about cognition. In the artificial intelligence approach, the researcher starts with a model of cognitive processes, uses this model to generate linguistic responses to linguistic inputs, and then evaluates the model in terms of the similarity of its performance with human performance in such contexts. Thus, the cognitive mapping approach requires a good representation of selected features of the assertions made by a person, while the artificial intelligence approach requires a good representation of the internal processes of a person.

(Continues...)

---

Excerpted from Structure of Decision by Robert Axelrod. Copyright © 1976 Princeton University Press. Excerpted by permission of PRINCETON UNIVERSITY PRESS.
All rights reserved. No part of this excerpt may be reproduced or reprinted without permission in writing from the publisher.
Excerpts are provided by Dial-A-Book Inc. solely for the personal use of visitors to this web site.

---

Table of Contents

• Frontmatter, pg. i
• PREFACE, pg. vii
• CONTENTS, pg. xi
• FIGURES, pg. xiii
• TABLES, pg. xv
• CHAPTER ONE. The Cognitive Mapping Approach to Decision Making, pg. 3
• CHAPTER TWO. Foreign Policy Formation Viewed Cognitively, pg. 18
• CHAPTER THREE. The Analysis of Cognitive Maps, pg. 55
• CHAPTER FOUR. Decision for Neoimperialism: The Deliberations of the British Eastern Committee in 1918, pg. 77
• CHAPTER FIVE. Complexity and the Presidency: Gouverneur Morris in the Constitutional Convention, pg. 96
• CHAPTER SIX. Explanation of the Unexpected: The Syrian Intervention in Jordan in 1970, pg. 113
• CHAPTER SEVEN. Strategy for the Energy Crisis: The Case of Commuter Transportation Policy, pg. 142
• CHAPTER EIGHT. Comparative Cognition: Politics of International Control of the Oceans, pg. 180
• CHAPTER NINE. Results, pg. 221
• CHAPTER TEN. Limitations, pg. 251
• CHAPTER ELEVEN. Projects, pg. 266
• APPENDIX ONE. The Documentary Coding Method, pg. 291
• APPENDIX TWO. The Questionnaire Method, pg. 333
• APPENDIX THREE. The Mathematics of Cognitive Maps, pg. 343
• APPENDIX FOUR. Simulation Techniques, pg. 349
• APPENDIX FIVE. Guide to Source Materials, pg. 360
• BIBLIOGRAPHY, pg. 373
• INDEX, pg. 395