The Bounds of Reason: Game Theory and the Unification of the Behavioral Sciences [Hardcover]

This book departs from a very good set of questions: How can it be that several different behavioral sciences - sociology, social psychology, economics, biology - all study human social behavior, yet have vastly different conceptual frameworks? And perhaps more importantly: is there a way to unify these frameworks?

Professor Herbert Gintis is uniquely qualified to tackle these questions: as any reader of his Amazon book reviews can see, he is very widely read in all the social sciences, and much of his own research is interdisciplinary. Gintis proposes that game theory - a mathematical framework for analyzing strategic interactions between individuals - can play the role of unifying framework for the social sciences. The first half of the book is dedicated to explaining the basic concepts of game theory, and how it applies to basic issues in human social behavior.

The second half of the book is dedicated to connecting game theory to the sociological concept of a social norm. A central point in Gintis' argument is the concept of correlated equilibrium. A correlated equilibrium augments the well-known Nash equilibrium by adding a correlating device. A correlating device - or choreographer as Gintis' calls it - essentially is a random variable with the distribution over the set of strategy profiles. The correlating device selects a strategy profile (one strategy for each player) and tells each player what to do according to this strategy profile. If it is optimal for each player to follow the advice of the choreographer given her beliefs about what the choreographer advised the other players, a correlating equilibrium exist. As an example of this one can think of a traffic light. When the traffic light tells you to drive, it is optimal to do so, because you know the traffic light simultaneously tells other people to hold still. Thus, the traffic light coordinates the actions on the intersection.

Gintis maintains that the correlating equilibrium is a better candidate to provide the technical underpinnings of the concept of a social norm than the Nash equilibrium. Gintis argues that the social norm functions as a correlating device, which assigns a particular action to everyone engaged in interaction. Actions are in equilibrium when it is in everybodies' interest to follow the social norm if they expect others to do so. The reason that the correlated equilibrium is more suited to describe a social norm than the Nash equilibrium lies in the conditions that underly both concepts. While the Nash equilibrium relies on the assumption that all players in the game have correct expectations about the actions of all the other players, the correlated equilibrium merely requires that people share a common prior belief about the actions recommended by the correlating device. If the correlating device is a social norm, these prior beliefs are induced by indicators that activate the social norm. How such prior belief can come to be shared is the subject of an entire chapter.

Professor Gintis should be praised for his commitment to methodological unification. Especially worthwhile in this respect is the last chapter, where he proposes several concepts that could form the shared theoretical background in all the social sciences. He also submits some sensible proposals, such as the use of the correlating equilibrium, and the comments on methodological individualism are provoking and stimulating.

However, for several reasons I fear that the book will have limited influence in actually bringing together scientific disciplines. First, its organization is at times mysterious, and never argued for. The connections between the chapters are often unclear. There are some parts that will not interest most readers, like the extensive treatment of certain paradoxes in epistemic game theory. This relates to the second problem: the book is rather technical in its exposition. This will make it hard for anyone who is not steeped in mathematics or does not already know game theory. And anyone who has taken the trouble to learn game theory up to the technical level that is required to read Gintis' book will probably be already convinced of its usefulness. Third, the book is lacking in good examples. Although the argument surrounding the correlated equilibrium is connected loosely to a story about flagging taxis, this raises more questions than it answers. For someone who want to convince sociologists that game theory can be connected to basic sociological ideas like norm following and role theories, this is a rather serious omission. Finally, there are some harsh dismissals of alternative methodologies in the social sciences which, although I tend to find them reasonable, are bound to bruise some egos. Altogether, it is hard to escape the impression that this is a book for game theorists, by a game theorist, which is puzzling given its stated aim.

In short, a stimulating and impressive book with sensible ideas. Unfortunately, I fear it lacks the didactic sensitivity to reach across scientific disciplines, and thereby does not quite fulfill its lofty aim.

I saw this book on a counter in a book store. I saw the quote on the front from Nobel-Prize-winning economist Vernon Smith likening it to the works of David Hume and Adam Smith---strong praise indeed. I was intrigued, so I picked up the book and leafed through it. I couldn't finish it at the book store so I bought it and took it home.

I can't do justice to this book in a sentence or two. Professor Gintis sets an ambitious goal for this book--reconciling the various behavioral sciences (economics, psychology, biology, and sociology) with their different, indeed conflicting, explanations of human behavior. He examines how the insights into human behavior given by game theory are complemented by understandings drawn from other social sciences.

Although I have little familiarity with game theory, I found the book relatively easy to follow. The writing is clear---with many short sentences rather than paragraph-long, pseudo-academic clouds of words. The author appears to have put considerable effort in to making the material accessible---he avoids difficult mathematics even if doing so slightly expands the needed explanations.

This book is both thought provoking and enjoyable. I strongly recommend it to any graduate student or professional in the social sciences. I plan to give copies as gifts to a few people.

It's only $25.20 if you buy it through Amazon rather than the $35 I paid in a bookstore.

================

An engineer who values economics

This book is an easy-to-understand survey of modern game theory with no more (and no less) mathematics than is necessary for that. It is not comprehensive. It covers only material that is logically consistent and grounded in observation. But it argues convincingly that that subset is all you should care about. Defining the boundary sometimes gets close to splitting hairs, an occupational hazard of game theory and any field trying to distill rigorous foundations from complex evidence and competing research traditions. Overall, however, the author lays down clear and sensible rules that exclude a lot of nonsense and organize the remaining material in simple and satisfying ways. Few specialists will agree exactly with the treatment, but I think almost everyone will find it reasonably good.

Therefore, if you want to learn modern game theory with a few days work, buy this book. Read it with pencil and paper to work out examples and exercises (it's not a text with problems at the end of each chapter but the author does occasionally leave proofs to be supplied by the reader). Use the Internet for some key references. It does not demand any special training in mathematics, nothing beyond eighth grade techniques, but the logic and set arguments can be very intricate. It requires attention and a precise mind to follow, but not calculus or any other form of complex computation.

On top of this, the author offers his own thoughts on how central concepts in game theory drawn from biology, anthropology, sociology, psychology and economics can be combined in a consistent framework, that can serve as a foundation for all five fields. Each field studies emergent properties that cannot be derived from the foundation, so each will need its own applied game theory. But there is value in building each on a foundation that is both logically consistent and consistent with observation. The rigor from that will reduce errors and illuminate cross-disciplinary insights. I think he's probably right here, but even if he isn't, it's a valuable way to compare insights across fields.

Any researcher concerned with behavior of living things who does not already know all the material in this book should read it.