Reviewed by
Bruce Edmonds
Centre for Policy Modelling, Manchester Metropolitan University
This book pulls no punches in its advocacy of Agent-Based Modelling (ABM) within Sociology. The first half of the preface, in particular, reads like a mini-manifesto for social simulation. In the first two paragraphs he writes: "In my view, sociology is not a branch of the humanities. Its task is to explain social puzzles by reducing reality to recurrent and simple patterns, as for any other rigorous science ... formalization, modelling and computer simulation can make the same difference here as for physics ..." (page ix)
and again:
"... formalized models that look at agent behaviour and interaction are essential elements for sociology..." (page ix).
Thus the book is uncompromising in claiming a central role for agent-based simulation. He then identifies a crisis in sociology, describing it in these terms:
"... paradigmatic fragmentation, parochial balkanization of research programs and schools, no epistemological and methodological consensus, low prestige of the discipline, low funds and poor influence on the public sphere." (page ix), noting that physicists have started to supplant sociologists in the understanding of social puzzles.
He then goes on to consider a different kind of sociology, one where computational models can occupy the "unexcluded middle position" between models that are "too complicated" and those that are "too abstract and oversimplified", building a bridge between the qualitative and the quantitative. This kind of sociology has a determinedly micro focus, with the macro features not being programmed in but rather emerging from the detailed interaction of heterogeneous agents within given social structures.
It does soften this initially stark message by stressing that the author does accept that ABM is not the only technique and should complement other sociological techniques, both qualitative and quantitative. Indeed it sees the potential of ABM to reduce balkanization by helping to "cross-fertilise" qualitative and quantitative approaches.
The first chapter introduces ABM: its history and main ideas. It provides a useful "map" of kinds of ABM with their different purposes, nature and uses, from abstract "Artificial Societies" to detailed "applied simulations", implicitly emphasising the importance (in the author’s view) of "Middle-range models".
The "meat" of the book is composed of two, very detailed, case studies. That is how agent-based modelling can help understand firstly "Cooperation, coordination and social norms" (Chapter 2) and then "Social Influence" (Chapter 3). In each of these the relevant literature documenting the sequence of ABM with their results and interpretations is covered in detail. Anyone wanting to understand the background and current state of the art in these areas would do well to read these.
This is followed by a chapter on methodology (Chapter 4) which does not restrict itself to abstract reflections but illustrates methodological issues with a detailed examination of example of the development and results of the relevant ABM involved in several disputes, thus vividly illustrating the need for replication and independent probing of simulation models.
The final, concluding, chapter seeks to draw out some lessons from all this ABM history, to look at some of the challenges these highlight, and suggest some ways forward. In this the nature and role of what the author means by "Mertonian middle theory" is made clear:
"... models should include details of behaviour and the social situation which means that we must translate abstract theories into empirically verified constructions" (page 167).
It also briefly looks at the prospects for helping elucidate social policy issues as well as looking at the potential place of an ABM-empowered sociology within the emerging transdisciplinary and competitive approaches that seek to understand different kinds of social phenomena.
The book has a rather odd and extensive appendix which is a printed transcript of the complete code of two simulation models which the author was involved in, which are discussed in the book. The code is copyrighted and it is not clear whether it has been made available under an open license. However the code is available from the author's website for the book (http://www.eco.unibs.it/computationalsociology/models.html).
The defects in the book are minor. The author takes a very optimistic view of the prospects for ABM in understanding social phenomena, concentrating on the upsides far more than the difficulties. I think that this will annoy some sociologists, especially those of a more qualitative bent, since the difficulties of the project will be manifestly obvious to them and these will not have been addressed to their satisfaction in this book. The book is enthused by optimism that there will be a convenient middle-range model to bride the gap between abstract theory and available empirics that is not too complex to be understood. However this is understandable in a book that is seeking to advocate an approach.
The book and its message will not be a surprise to JASSS readers, but it does make a cogent and coherent argument for ABM that is especially suited to those coming from sociology. It puts the models and their results firmly within a sociological context and argument and does not get lost in the technicalities of the models. It will not convince all sociologists, indeed some will see it as confirmation of their suspicions, but in any case it will inform and demonstrate how ABM can give insights into social phenomena. The chapters that introduce ABM and its methodology are useful, but it might be the two careful and detailed case studies that have the most impact. This book should be inserted into all sociological libraries as a vanguard for the rest of us - if it not torn to shreds by enraged sociologists it will very usefully inform them. Newcomers to ABM and even old hands, but especially those who have to survive within sociology, will find it a very valuable asset.
Return to Contents
of this issue