© Copyright JASSS

JASSS logo ------

Social Emergence: Societies as Complex Systems

Sawyer, R. Keith
Cambridge University Press: Cambridge, 2005
ISBN 0521844649 (hb), 0521606373 (pb)

Order this book

Reviewed by Edmund Chattoe
Department of Sociology
University of Oxford

Cover of book As a simulator who has spent sufficient time amongst sociologists to know something about them, I enjoyed reading this book. It is clearly written and describes challenging theories in an accessible way. I also found it thought provoking for my own research. However, I am rather unconvinced that it will "contribute to resolutions of long-standing unresolved issues in sociology and provides methodologies that are of immediate use to sociologists…." (p. 2).

The problem lies as much in the existing culture of sociology as in the book itself. A significant portion of the book is devoted to social theory, specifically to describing and evaluating competing accounts of emergence and linking these to their historical antecedents. Unfortunately, the theories considered illustrate why the book is unlikely to reconcile these competing accounts. It becomes painfully clear that the subject matter of the social theory discussed is not the social world or the methodology of social science but the work of other social theorists. Even in philosophy, it is common to organise analysis around recognised examples or "problem cases" (albeit heavily stylised). Here, concrete examples are in extremely short supply and, when they do occur, are often "sketches" from the physical sciences rather than detailed accounts of social settings (the proverbial gas molecules and the rules of chess).

As a result, this reader found it almost impossible to decide whether the competing theories had any distinctively different implications for practical sociology let alone how one would establish which was most useful or even "true" in a concrete context. It seems that this style of social theory vigorously avoids the danger of "another good story ruined by an eye witness". (Appropriately, this quote is variously attributed to Alben W. Barkley and W. C. Fields!) From the perspective of the social theory "language game", Sawyer has made a relatively analytical and jargon-free contribution but as far as empirical sociology goes, more social theory (even if it is clear and thoughtful) is in danger of adding to the problem rather than providing the solution.

At the other extreme within the discipline, empirical sociologists simply refuse to lose much sleep over emergence and will need quite a lot of convincing to fix something that they don't regard as broken. Furthermore, simulation appears (at least in principle) to tackle many of the more substantive problems troubling the relationship between different research methods. Social statisticians look for methodologically valid associations in aggregate numerical data but are challenged for not really explaining those associations. Qualitative researchers describe the richness of individual experience and social settings but often either do not (or interestingly do not wish to) generalise from these accounts and build "middle range" theories. In consequence, they are often accused of journalism or story telling. In simulation, the "Gilbert and Troitzsch box" (Gilbert and Troitzsch 2005, p. 16-18) suggests how a simulated system (abstracted from a real system) can be used to generate simulated data that can then be compared with real data offering the potential for falsification. In this methodology, statistical analysis is no longer "ungrounded" at the individual level and ethnography (or at least insights from it suitable to design agent based simulations, see for example Agar 2003, 2005) is no longer unable to generalise because of the potential complexity of social interaction. None of this appears to hinge (at least as far as I can see based on the arguments presented by Sawyer) on what stance we take about emergence. Both simulations implementing strong methodological individualism and those that involve agents with shared representations of norms and social structures seem to sit equally comfortably within this methodology.

In terms of clarifying the concept of emergence, the most promising part of the book is also easily the thinnest and this seems like a badly missed opportunity. By far the most concrete discussions of social emergence in its various forms are found in the descriptions of existing simulations in chapters 8 and 9. (This provides further indirect evidence for the view that social theory is foundering because it has nothing substantive to get its teeth into. Even clearly specified toy simulations – like examples or "problem cases" – allow more to be said clearly than "pure" theory.) Given this advantage of simulations, it is rather surprising that Sawyer did not attempt to develop his own model based around recognisable qualitative features of emergence. Agent based simulation seems like precisely the technique to explore whether various kinds of shared mental representations (which may be independently located in material objects like books, the media and computer software) can or cannot be usefully considered as reducible to individual mental content. (My view is that once individuals reason based on sufficiently complex social facts, it may remain strictly true that individual mental content is explanatory and causal but it will no longer be possible to develop a workable methodologically individualist account for the creation and transmission of those social facts.) The weakness of not developing an appropriate simulation of emergent phenomena is rhetorical as well as scientific. Empirical sociologists who are already skilled in methods of statistics and ethnography are unlikely to be "sold" a different method by someone who merely tells them about it. In a sense, Sawyer is not following his own advice about the value of simulation as a theory building tool in the emphasis of his book and this seems likely to provoke scepticism. (From my own experience, sociologists can be pretty sceptical even when presented with output from a working simulation in a domain that they have already defined as properly sociological.)

In conclusion, then, I suspect simulators with an interest in emergence will browse this book for interesting ideas to inform their model building. Social theorists will probably add it gleefully (and with a carelessness it does not deserve) to their ever-expanding soup of terminology but empirical sociology will, unfortunately, ignore it.