This new book by Jon Elster purports to be ‘an elementary, informal, and personal presentation of ideas that have’, he believes, ‘considerable potential for illuminating social behaviour’ (p. ix). Elster's original intention was to update Nuts and Bolts for the Social Sciences, an attractive little book published in 1989. What turned out is this volume, three times as long, covering more topics and entering into more details.

Let me give an idea of its content, leaving aside for the moment Part 1 and the Conclusion. Part 2 is devoted to ‘the mind’, with chapters on motivations, self-interest and altruism, myopia and foresight (with a good, accessible, introduction to hyperbolic discounting), beliefs, and emotions, Part 3 to ‘action’, with chapters on desires and opportunities, persons and situations, rational choice, rationality and behaviour, responding to irrationality, and ‘some implications for textual interpretation’. I found the development on ‘persons and situations’ particularly interesting. The main idea (not Elster's, originally) is that individual behaviour is the result of situation more than of character. The fact that folk psychology suggests otherwise and that, therefore, we also tend to think otherwise when interacting with other individuals has behavioural consequences to which Elster returns later in the book. The chapter on textual interpretation is about assessing consistency in novels – perhaps a digression, but quite fascinating as well, at least if one is a reader of fiction. Entitled ‘lessons from natural sciences’, Part 4 includes three brief chapters on physiology and neuroscience, explanation by consequences and natural selection, selection and human behaviour. The following section, Part 5, entitled ‘Interaction’, is divided into nine chapters. There, a brief and relatively disappointing chapter on unintended consequences is followed by two chapters on game theory, followed themselves by individual chapters on trust, social norms, collective belief formation, collective action, collective decision making and ‘organisations and institutions’. To avoid being misled by the titles, one must stress that, in that section also, the main reliance is on psychology and history.

Even when analyses and arguments seem familiar, they are typically accompanied by new (or forgotten) stimulating illustrations, remarks and references. This remains true whether Elster summarizes analyses he developed himself in previous work (rationality, sour grapes, emotions, social norms, constitutional conventions), presents path-breaking analyses developed by others (hyperbolic discounting, situation versus character) or exposes in a reader-friendly way pieces of standard theory (utility, Bayesian updating, game theory). Thus, even if, as advertised by the publisher, Explaining Social Behavior may be used as a textbook, one may also decide to read it in a relaxed way, skimming through some parts, reflecting on others, simply for the general intellectual enjoyment and stimulation the book provides, a bit like reading and re-reading the authors that Elster himself reads and (probably) re-reads, Proust, Stendhal and Tocqueville for instance.

For readers of this Journal, the most relevant aspect of the book is different. As the author stresses, it ‘has a methodological and philosophical slant not usual in introductory-level presentations’ (p. ix). In fact, three methodological chapters – entitled, respectively, explanation, mechanisms, and interpretation – make up Part 1, and the 22-page conclusion is entitled ‘Is social science possible?’ In addition, there are methodological and philosophical assertions at the beginning of each part and in most chapters. The normal way to discuss the methodological dimension of the book would consist in focusing on the methodological views expressed by Elster, considering that the more voluminous – and in some respects more attractive – analyses belonging to psychology or social science are only supportive or illustrative of these views. But this would lead to the formulation of a systematic assessment of the said views, a task that I want to eschew for several reasons. On many issues I simply do not have the qualification. For instance, although I am not fully convinced by the discussion on causality in Chapter 1, I am unable to discuss the arguments competently. I share Elster's interest for mechanisms, laboratory experiments, psychological hypotheses, history and novels (I am less sure about the French moralists, who bored me to death when I was a child). But this does not preclude finding some merit in covering-law forms of explanations (broadly understood), (retro)predictive power, statistical tests, rational choice as a research strategy and, last but not least, modelling. There is no space here to elaborate the differences between these positions.

In the Conclusion, Elster, who calls his favoured approach ‘qualitative social science’, draws from the substantive analyses developed in the book a list of the achievements of that approach, which he contrasts with two others, ‘soft’ and ‘quantitative social science’ – manifestations, he claims, of ‘soft’ and ‘hard obscurantism’ respectively. Soft social science (consisting of postmodernism, etc.) is dealt with in less than a page. The criticism of measurement and data analysis, as part of ‘hard social science’, is actually a set of warnings against well-known pitfalls. Elster's main target, thus, is what he calls ‘mathematical’ modelling. When he turns to that activity, economics (only implicitly included in social science up to that point) makes its open entry as an object of criticism and insinuations expressed in surprisingly harsh terms (p. 461). The consolation offered economists in fine is that– although this ‘is not a choice we have to make’ – ‘the pathologies of hard obscurantism are to be preferred to those of soft obscurantism’ (p. 465).

Earlier in the book, Elster had introduced the chapter on game theory with the assertion that its invention may ‘come to be seen as the most important single advance of the social sciences in the twentieth century’ (p. 312). For that reason and others it seems clear that what Elster really objects to cannot be economic or formal theory in toto but something like excessive formalization or, more interestingly, exclusive reliance on modelling. The complaint against excessive formalization is commonplace. So are all the arguments formulated by Elster against economics as a science as well as the ten points he makes against various forms of modelling. As he himself suggests, it is easy to predict the replies. Thus, a systematic assessment of the explicit part of Elster's methodological position might be somewhat tedious – an additional and clearly decisive reason not to undertake it here. An exception is the (implicit) resentment against exclusive reliance on mathematical modelling, with which I sympathize. Modelling is not always fruitful and it need not always be formal (as explained below).

Instead of discussing the methodology exposed in the book, understood now as a way to defend the type of substantial analyses presented in the non-methodological chapters, we may take the intrinsic interest of these analyses as fact and raise the question: how may that fact be analysed in methodological terms? Elster's methodological chapters can be read then as offering one answer. But this is not the only possible one. We may seek other ways to account methodologically for the analyses or complement Elster's own account. We are fully entitled to do this because the substantive creations of scientists can be separated from the methodological interpretation they or others give them (the same is true of works or art and their interpretation).

The main ingredient in Elster's ‘qualitative’ method is a focus on mechanisms – in most cases with a large psychological component. But for him, hypothesized mechanisms – or more generally theoretical hypotheses – are to be judged exclusively on the basis of their contribution to explanations (their truth being only a necessary condition for that, and prediction being out of reach). This may account for the conflation (p. 17) of the hypothetico-deductive method (which concerns evaluation) and the deductive-nomological model (which concerns explanation). The quasi-impossibility of satisfying in natural settings the demanding criteria that Elster sets for a correct explanation inspires his pessimism about what social science can achieve. As he concedes with regard to his answers, listed in the Conclusion, to the questions or puzzles he had listed in the Introduction, in most cases ‘they satisfy only the minimal condition of implying the explanandum’. Elster's pessimism about social science is also reflected in his definition of mechanisms: ‘frequently occurring and easily recognizable causal patterns that are triggered under generally unknown conditions with indeterminate consequences’ (p. 36).

I will make three points:

• – mechanisms may be a legitimate object of inquiry for themselves, independently of their role in explanation;

• – mechanisms may be worthy of interest because of their role in arguments;

• – mechanisms may find their place as components of models.

The last point is particularly important for economics.

Mechanisms are typically characterized by reference to their direct effect. The ‘sour grapes’ mechanism is defined as the adaptation of preferences to possibilities. Trivially, under some conditions of isolation or abstraction, and thus, typically in an artificial context such as that of a laboratory, once ‘triggered’ it explains – and, one may argue, predicts – that preferences are adapted to possibilities. Similarly, once triggered, the ‘grass is greener’ mechanism explains and predicts, also trivially given its definition, that preferences are ‘counter-adapted’ to possibilities. Even in artificial settings it may be a bit difficult to predict, or even explain ex post, the occurrence of triggering but once it is established that one of the mechanisms is at work, the direct consequences should be straightforward. In the definition of mechanisms cited above, it is clear that Elster does not have this context in mind. He places himself in a natural setting, where it may be difficult not only to know what exactly triggers a mechanism but also, because many other forces may operate at the same time, what are the outcomes. Hence his pessimism about the capacity of a mechanism-based social science not only to predict, as he stresses at the outset, but also to explain as he more or less concedes also.

Now, motivations, preferences, beliefs, regularities, processes, capacities, dispositions, and mechanisms are features of the real world (although, as we will see, they can also be introduced in models). As such, they are interesting and a legitimate object of inquiry independently of their role in the explanation of phenomena. In particular, discovering a mechanism or a new property of a known mechanism is by itself a valuable achievement, even if, in natural settings, the explanatory implications are for the moment minimal or the relevance unascertainable. The discovery of hyperbolic discounting as a mechanism is an achievement that contributes to explaining observed phenomena such as procrastination or lack of will power. But its intrinsic properties are interesting and worth studying, as illustrated by the fact that findings in physiology and neuroscience have inspired the spin-off of a variant: quasi-hyperbolic discounting or salience of the present.

Although Elster bonds together mechanisms and explanations, the way he dissects the former reflects a curiosity in them independent of their use as ingredients in the latter (even if he also displays a pervasive concern with such use). Manifestations of curiosity are also observable in the way Elster discusses preferences, motivations, emotions, and norms. His fine distinctions and elaborate taxonomies make one think of the skill of an entomologist, whose interest in insects for themselves is not doubted. It should be noted also that what we get from the book under review, as from previous work by Elster, is not only a large number of individual emotions, motivations, and mechanisms but a partial organization of their relationship which is quite stimulating in itself. However, the ‘intrinsic interest’ perspective stressed here and Elster's definition of mechanisms do not fit. Under the said perspective, mechanisms do not have to be frequently occurring and easily recognizable. What counts is that they exist and that as many as possible of their properties be known. In one case at least, Elster denies that a psychological mechanism exists although other authors assert that is does (pp. 73–4). Mechanisms are investigated in various ways. A mechanism may be identified and studied in the laboratory and the way it operates in the open universe in general or in particular historical circumstances analysed by other means – the transition from the confined atmosphere of the laboratory to tempestuous natural settings being often messy. Anyhow, perhaps the main consequence of the possibility of an intrinsic interest perspective is that a failure of a mechanism to contribute to explanations in a fully ascertainable way becomes less troublesome. Lack of confidence in the asserted mechanism-based explanation of a phenomenon does not necessarily translate into doubts about the mechanism itself.

My second point also starts from a self-mutilation that Elster imposes on the relevance of his own analyses and of the qualitative social science that he favours. For him, explaining the possibility of a phenomenon is interesting only as a first step in the construction of a full-blown explanation (p. 25). This neglects the fact that social science, and especially mainstream economics, is first of all a critical enterprise. It is grafted on the multifaceted universe of beliefs, ideas, and theories held or entertained by all segments of the public, including government, about human behaviour, social phenomena, and their explanation. Showing that there might be plausible mechanisms accounting for what seems mysterious, that what is believed to be impossible may be possible after all, and that what is believed to be possible is in reality impossible (here, we approach the traditional nomological role ascribed to scientists and in fact a major contribution of economics) are achievements that justify the existence of a social science even if it could not provide the ‘successful’ explanations on which Elster focuses his methodological attention. He also claims that ‘we should be cutting ourselves off from many insights if we ignored the mechanisms suggested by philosophy, fiction, plays, and poetry’ and that it is a loss ‘if we neglect twenty-five centuries of reflection about mind, action, and interaction in favour of the last one hundred years or the last ten’ (p. x). But we misunderstand and underestimate the wisdom of authors such as Montaigne and the 17th century French moralists cited by Elster if we neglect the fact that they often viewed their analyses or assertions, including about mechanisms, as arguments in an on-going debate with the rest of society, in particular with the ‘received’ wisdom prevailing in it. This applies also to the way mechanisms understood in artificial settings can serve as arguments in natural ones. And it may also apply to mechanisms in models, to which I turn now.

Explaining Social Behavior is about individual behaviour in social contexts rather than about social phenomena in general. I am not sure that Elster accepts that distinction but I think that it is an important one to understand the purpose of economics and perhaps also of the other social sciences. Given the real subject of the book it is natural that it stresses psychology and discusses only relatively simple (even if subtle) interactions between individuals. This may be enough to explain some kinds of large-scale phenomena, even involving millions of individuals. As a rule, however, the mechanisms that underlie social phenomena, in particular in the economic sphere, involve more complex interactions, and moreover interactions in which quantitative constraints play a major role. This makes recourse to models indispensable. Economics has thrived since the eighteenth century on the conviction that the neglect of hidden processes, interactions, and mechanisms causes major errors (including regarding policy making) and that, because of their complexity the only way even simply to think about these hidden processes, interactions and mechanisms is by constructing models.

Now, philosophy of science is still as uncertain about the nature of models as it is about that of mechanisms. One interpretation of models is as fictional worlds or societies, described with the help of mathematical models or in plain language, and in which one may introduce (or eventually derive or discover) the same entities and furniture (preferences, beliefs, capacities, constraints, activities, mechanisms) as in reality. Usually though, models are constructed more economically so as to be tractable and allow deduction. This is justifiable because models need not represent reality. For a model to be useful, one must be able to formulate a conjecture that some things in reality are or may be as some things in the model. Following that line, we may say that Elster, as many commentators, misunderstands the ‘as if’ supposition (pp. 31, 464). It should not be stated that individuals in a model are assumed to behave ‘as if’ they were maximizing utility (which indeed verges on the absurd) but that something in reality is or happens as does something in a given model in which individuals are set to maximize utility. The something mentioned here may be a mechanism, a process, a relation, a property, a result, an event.

Under this interpretation, models are fictions and instruments by definition. But their use may or may not reflect a philosophical position of instrumentalism. Sometimes, the hypothesis or conjecture designating something in reality as being like something in the model is limited to the prediction of a relationship between observables, no similitude being assumed between the behaviours or mechanisms that underlie the relationship, in which case one may want to speak of instrumentalism. In other cases, we are curious about the reality under the surface. Then, although individual behaviour is not assumed to be perfectly similar in the model and in reality, the conjecture purports to elucidate the working of some mechanism or interaction in reality by referring to a similar mechanism or interaction in the model. In such cases, I don't think that speaking of instrumentalism is justified. Because of its fruitfulness, rational choice modelling may be justified along that line, but so may also be the use of models in which rational choice is qualified or even given no role at all.

The purpose of the foregoing digression is to suggest that, as ‘nuts and bolts’, a comfortable habitat for mainly psychological mechanisms might be located not only in the real world but also in the models that economists deem necessary to discuss some social phenomena. This would have been impossible two decades ago because all mainstream economists were still shy of departing from strict rational choice as the main feature of their constructs. This is not true anymore. In Bénabou and Tirole (Reference Bénabou and Tirole2006), for instance, several of the psychological mechanisms presented by Elster – especially dissonance reduction and hyperbolic discounting (in the form of salience of the present) – play a major role to account for a number of complementary differences between the United States and Europe. Thus Explaining Social Behavior should be recommended also to young mathematical social scientists in search of inspiration.