Reviewed by
Peter Davis
COMPASS, University of Auckland
This is a volume about analytical sociology edited by Pierre Demeulenaere of the Sorbonne and published in paperback by Cambridge. This is a book of obvious interest to sociologists; as a quote from Peter Bearman of Columbia on the cover states, "analytical sociology is - simply - good sociology". A strong claim indeed for that community, but why should this be of interest to the simulation crowd?Analytical sociology seeks to develop explanations of social phenomena by identifying the mechanisms that generate those events or outcomes. It does this by breaking down social processes into their key components and then showing how those components interact to produce the outcomes of interest. These are called social mechanisms. Hence the full name of the book - analytical sociology and social mechanisms.
Two examples are fully worked in separate chapters in the book, both pitched at the neighbourhood level of analysis: the role of social interaction - peer groups in this case - in explaining variable levels of youth unemployment across different neighbourhoods (Youth Unemployment: A Self-Reinforcing Process?, by Yvonne Åberg and Peter Hedström), and the part that selection and sorting mechanisms play in shaping and reproducing a durable and racially and socio-economically ordered residential stratification system (Neighborhood Effects, Causal Mechanisms and the Social Structure of the City, by Robert J. Sampson).
The relevance to simulation is developed in two further chapters. Thus, in a contribution by Michael W. Macy and colleagues (Social Mechanisms and Generative Explanations: Computational Models with Double Agents, by Michael W. Macy, Damon Centola, Andreas Flache, Arnout Van de Rijt and Robb Willer) the authors use phrases such as "generative explanations", "social life from the bottom up" and "the microfoundations of social complexity". The promise is of mimicking and reproducing social processes through the creative use of humanoid-like agents programmed to follow simple rules of behaviour in forging social events and outcomes - from the "bottom up".
A second dimension to the simulation task is developed by Gianluca Manzo in a chapter on relative deprivation in silico, i.e., in a virtual environment (Relative Deprivation In Silico: Agent-Based Models and Causality in Analytical Sociology, by Gianluca Manzo). In essence, not only can we try and mimic social processes through agent-based computational techniques, but we can also test various counterfactual propositions, a feat rarely possible in standard non-experimental social science research. And this relates to current theories about making causal claims, the so-called potential outcomes framework which argues that an effective claim is one that can meet the test of plausibly demonstrating "what might otherwise have been, had circumstances been different". Thus, computational techniques provide tools to test "potential outcomes" in silico because we are able to try out different configurations of elements in key social mechanisms to see how well they reproduce events and outcomes, with the ultimate test being validation against real empirical benchmarks.
The book has an introduction and thirteen chapters distributed evenly across three sections: actions and mechanisms, mechanisms and causality, and approaches to mechanisms (the four chapters already alluded to - two on empirical application at the neighbourhood level and two on the use of simulation techniques). Aside from the empirical exemplars, the predominant emphasis in these chapters is programmatic, philosophical and technical (e.g., Ordinary Rationality: The Core of Analytical Sociology, by Raymond Boudon, Indeterminacy of Emotional Mechanisms, by Jon Elster or Generative Process Model Building, by Thomas J. Fararo). Indeed, one chapter was so technical that it eluded this methodologically-experienced reader (i.e., chapter 6 by Peter Abell), and another chapter contained sixteen pretty indigestible pages of graphs (i.e., chapter 13 by Gianluca Manzo)! A concluding chapter to bring the threads together would have assisted in removing the impression that this is a book of readings that do not really hang together, albeit organised under the motivational banner of the analytical sociology "movement".
Two chapters make contributions that are of substantive interest in framing analytical sociology (i.e., Conversation and Mechanism: Emergence in Creative Groups, by Keith Sawyer and Neighborhood Effects, Causal Mechanisms and the Social Structure of the City, by Robert J. Sampson). Thus, Keith Sawyer argues that emergence is a uniquely social - and hence sociological - phenomenon that cannot necessarily be reduced to constituent parts, and indeed is of special interest for that reason. And Robert J. Sampson makes the point that higher-order processes that induce structure at the neighbourhood level, together with causal processes of institutional and historical dimensions, require "a more flexible conception of causality than that offered by individual experiments and even, dare I say, methodological individualism".
As a sociologist, I would like nothing better than to open, say, the sociology equivalent of the Samuelson economics student text book and find a pleasingly-structured analytical framework applicable across every dimension of society. But we are not there yet. And perhaps we never will be. At present the "whole" of analytical sociology is somewhat less than the sum of its parts - a criticism often levelled at the analytical approach itself! - but there is plenty of work here for those members of the simulation community willing to come more than half way across the divide between the technical/computational and substantive/social science wings of the simulation paradigm.
Return to Contents
of this issue