Reviewed by
Edmund Chattoe-Brown
Department of Sociology, University of Leicester
The most interesting thing about this book is what it doesn't do. It is a collection of articles forming a special issue of the Journal of Mathematical Sociology . It presents approaches to the "Micro-Macro Link" (MML) featuring experimental (1), analytical (2), narrative (2) and simulation based (3) contributions all of which are interesting and well written. However, reading the whole book, it soon emerges that each approach interprets MML in its own way even in the light of the other contributions. The book thus has limited synthetic value. (The introduction has interesting aspects - for example discussion of the "explanatory sociology" research programme that has probably been neglected since very few of the participants published in English - but tends just to describe rather than draw together the diverse strands of relevant research.) As an example of the book tending to ignore its own findings, the chapter based on experimental economics (by Gächter and Thöni) treats the "macro" level as a relatively straightforward aggregate of individual behaviour. This assumption (while common in economics) contrasts sharply with the non-linear and counter-intuitive aggregate outcomes of the Schelling model (discussed by Fossett). In a similar vein, where the non-simulation chapters mention simulation at all, they rehearse the (rather outdated) view that its role is just to deal with mathematical systems that cannot be tackled analytically even though it is very clear that this is not (for good or ill) what the chapters by Fossett (and by Flache and Macy and Helbing et al.) are doing.
In fact the "generative approach" to simulation (discussed by Gilbert and Troitzsch and by Epstein) gives the simplest way of seeing how each method is limited and fails to engage with the opportunities presented by the others. The experimental paper produces very interesting data but struggles to explore the dynamic decision and interaction processes that might have given rise to it (getting only as far as behavioural "types" and largely controlled laboratory interaction). By contrast, the simulation papers all suffer from ambiguity regarding what specific facts about the world they are trying to explain. In particular, the tendency seems to be to produce somewhat plausible generic properties like "conformity/polarisation" or "cooperation/defection" rather than specific patterns based on "standard" social science data (like surveys or field research). Thus, in the same book, Helbing et al. can establish robust cooperation in a stylised world while Gächter and Thöni explicate the "inevitability" of the Tragedy of the Commons. Unfortunately, the development of a wide range of possibilities doesn't seem to resolve the issue of explaining how the world really is in particular contexts. There still seems to be a gap between the goals that modellers set themselves and what those with practical problems might expect them to be trying to achieve.
The two least convincing chapters are those by Opp (which really struggles when compared with the methods/data based chapters to say anything substantive especially as it spends some time rehearsing questionable defences of Rational Choice as a social science research paradigm - surely by now this disagreement is either outmoded or sterile?) and by van de Rijt, which is odd in building a detailed analytical model based on what is basically a "folk saying" (namely "the friend of my friend is my friend") albeit one that has a long tradition in research. (This seems to be an area, like the Prisoner's Dilemma where the research agenda has taken on a life of its own.) Even van de Rijt admits that the data for this behavioural pattern is "equivocal" (p. 101) but doesn't seem disposed to explore what kind of model could be built from more unequivocal data.
Thus, taken as a whole, the book richly illustrates techniques and data suitable for representing "agent models", interaction processes (perhaps under varying degrees of experimental control) and the potential for generating "rich" (rather than simple aggregate) macro patterns. However, the chapters in the book, far from taking these things together (as one might have hoped in a single volume under editorial control), have a strong tendency to take them apart (and also to isolate them from standard forms of pre-existing social science data). Thus we see no contribution that starts from data about individual behaviour (and a "real" environment) develops a theory than can mirror that system and then sees if it produces the "right" macro phenomena. Each paper misses out at least one aspect of this potentially falsifiable "generative" methodology and some miss out several. Thus, while the book makes an interesting and well organised read (particularly for those new to the debates) I am not sure how far it has advanced a progressive research strategy for understanding the MML. It may be that this collection was better suited to be a set of journal articles than a book.
Return to Contents
of this issue