Reviewed by
Juliette Rouchier
Groupement de Recherche en Economie Quantitative d'Aix-Marseille
The aim of the book is to help all behavioural scientists (from anthropologists to psychologists) to understand game-theory better. In particular, it helps to see in which circumstances a game-theory framework should be used and what are its main limits. Although Gintis strongly believes that a game-theory framework can help to complement social sciences, he does not pretend that it can cover all the spectrum of social phenomena of interest and suggests to take it as a complementary approach to others. In particular, he declares that "future advances in economics will require that model-building dialogue with empirical testing, behavioural data-gathering, and agent-based models". As a matter of fact, Herbert Gintis is one of the few well-known economists who has developed agent-based simulation models in his past research.Therefore, there are two aspects that are both thoroughly developed in this book. Firstly, since it provides a detailed description of all the different dimensions that are involved in decision theory and game theory, the book at some point looks like a (not necessarily basic) course in classical economics approaches, including many possible applications in the analysis of social facts. The second aspect is the explanation of the limits of the described approach. Gintis offers new insights from economics that expand the understanding of sociality, such as behavioural economics or evolutionary economics.
A social scientist who is not good at mathematics could be a bit afraid of the book at first sight, since many parts are full of formalisms and proofs that are far from basics. However, Gintis interprets in a very intuitive way most of the models and theorems he presents and most chapter introductions might help those who would like to understand the main concepts of game-theory without looking into technical details.
Decision theory and Game theory are explained and analysed following two angles. Firstly, it is shown that there are many known contradictions between assumptions in the basic theory and the reality of human behaviour. Real humans are not capable of following the complex reasoning that is needed to solve most of the problems (e.g., perception of probabilities is a problem and people cannot do backward induction properly) or display preferences that are not consistent with the theory (time-inconsistency). To explain this, Gintis goes first through the basics of decision theory and game-theory and then provides results from lab experiments drawn by the recent discoveries of behavioural economics. In doing so, the book can teach the classical rational hypothesis and show many features of actual behaviour of humans, as well as the heterogeneity among individuals.
Secondly, Gintis shows that even if one would like to follow a game-theory framework to analyze interactive strategic choices, the situations that agents would face would never generate social life as we know it. Starting from chapter 4, he elaborates on certain limits that game-theory encounters while dealing with sociality. For example, Common Knowledge of Rationality (CKR), where each agent knows that the other agent knows and so on, is to be assumed ex ante and cannot be a result of the game situation itself. Backward induction, which enables agent to choose the best strategy, is mostly in contradiction with Common Knowledge since it can lead to misinterpretation of other's intention. When several strategies are exactly equivalent in terms of pay-off, there is no reason why an gent would choose the strategy that benefits society. All these elements (and more) imply that one cannot get rid of some decisions or constraints that are defined from outside the game - norms or a way to play to get a commonly agreeable situation. Again, the justifications are a mix of formal proofs and explanation on how to interpret the formalism and the text can be a bit difficult for those who never use mathematical proof in their everyday life.
With this book, Gintis wants to promote the unification of the behavioural sciences, which 'include economics, anthropology, sociology, psychology and political science, as well as biology insofar as it deals with animal and human behaviour.', bringing each discipline toward accepting some common standards, among which game-theory is a good candidate. But economists, on the other hand, should understand that they cannot explain many phenomena with this approach, and the book nicely integrate these two aspects. As such, it is a good reading, in which all types of readers can find interesting information and skip too complicated mathematical proofs.
In conclusion, I only felt a bit uncomfortable with one aspect of the Gintis' excursus: by trying to justify the existence of sociality in the perspective of the strategic rationality, he refers a lot to evolutionary dynamics, and considers that processes that survived evolution are "better" than those who disappeared, which is for me a very normative way of seeing evolution, not necessarily appropriate in itself. Apart from this point of view (which is common to most evolutionary economists), I think the book is worth reading and shows a positive change in the desire of economists to really communicate and interact with colleagues of other disciplines.
Return to Contents
of this issue
© Copyright Journal of Artificial Societies and Social Simulation, 2010