© Copyright JASSS

logo --------

Social Science Microsimulation

Edited by Klaus G. Troitzsch, Ulrich Mueller, G. Nigel Gilbert and Jim E. Doran
Berlin: Springer-Verlag
Cloth: ISBN 3-340-61572-5

Order this book


Reviewed by
Brendan Halpin
Institute for Social and Economic Research, University of Essex.

Cover of book

This is a volume of what are effectively proceedings for a meeting of social simulators, held at the Schloß Dagstuhl in 1995. As such, it constitutes prima facie evidence for the stimulating nature of that meeting, in the broad range and vitality of the papers, which, although naturally variable in quality, contain a number of statements about the "state of the art" in various techniques for the simulation of social phenomena. As well as the twenty substantive papers, the plenary paper by Gilbert (Simulation as a Research Strategy, pp. 448-454) and the two summarised group discussions (Environments and Languages to Support Social Simulation, Computer Simulation and Social Sciences: On the Future of a Difficult Relation) give the feeling of a lively and timely meeting.

The range of papers is broad, even diverse, to the extent of risking a loss of focus. Indeed, in each of the four thematic sections of the book, there are papers with only the most tenuous link to the heading of the section. (In the extreme, one paper - while interesting - contains only a passing reference to simulation.) Nevertheless, on the whole the result is a coherent and interesting collection.

The main strength of the book is in pulling together up-to-date work in a number of distinct and important areas in social science simulation. (As the title declares, this is simulation as micro-simulation. There is nary a whiff of system dynamics here.) The areas represented include "micro-analytic micro-simulation" (tax-benefit models and so on), cellular automata, game theory, multi-level simulation and distributed artificial intelligence. Furthermore, the bulk of the papers have a genuinely social focus, rather than merely involving multi-agent simulation in a nominally social context. The paper of which this is least true is Molnár's model of pedestrian flow (A Microsimulation Tool for Social Force Models, pp. 155-170): an entirely sound paper in itself but one in which "social forces" are simply the mutual effects which pedestrians have on each others' movements. However, in the main the substance of the volume is rather more clearly social scientific.

Ask an economist about micro-simulation, and they will usually respond exclusively by reference to survey-calibrated tax-benefit simulation models - the microeconomic equivalents of the "treasury model". On the other hand, this field seems to be regarded from within the simulation community as a sort of poor relation, a relatively mechanical application of expert knowledge to survey data. Therefore it is good to see a number of papers here dealing with the methodology of what has been labelled "micro-analytic micro-simulation" (MASM). From the point of view of social science simulation, this is an important methodology, because it deals directly with social processes at the level at whichthey are conventionally analysed and measured. It represents one extreme, perhaps, in the trade-off between simulation flexibility and verisimilitude, but it is a trade-off that all simulations are obliged to make. Thus the papers by Heike et al. (A Comparison of 4GL and an Object-oriented Approach in a Micro Macro Simulation, pp. 3-32) and Merz (MICSIM: Concept, Developments, and Applications of a PC Microsimulation Model for Research and Teaching, pp. 33-65) are very welcome in presenting this methodology to a simulation audience strictly as a form of simulation. (The two papers present respectively an insight into the sophisticated database-manipulation "guts" of the micro-simulation process and an outline of a MASM program designed for the benefit of researchers who actually want to use existing simulations.)

The second section of the book is devoted to another domain, that of multilevel simulation. This section has the advantage that the papers cohere well and complement each other, presenting an overview of the multilevel simulation approach by Troitzsch (Multilevel Simulation, pp. 107-122), a detailed discussion of the MIMOSE package by its author, Möhring (Social Science Multilevel Simulation with MIMOSE, pp. 123-137) and an application to a real problem: Saam's modelling of the patterns of military intervention in Thailand (Multilevel Modelling with MIMOSE: Experience from a Social Science Application, pp. 138-154). This three-legged treatment of the approach is very useful to readers interested in MIMOSE or other multi-level approaches, providing a real opportunity to consider their value.

Cellular automata (CA) and game theory, as often happens, are taken together in the third section of the book. Here we find a strong argument for the relevance of CA in social simulation provided by Hegselmann, incorporating interesting history and some useful modelling (Understanding Social Dynamics: The Cellular Automata Approach, pp. 282-306). Hegselmann makes a persuasive argument for CA on a number of grounds (the importance of time and space, the bottom up nature of social action and so on) which make one reconsider the value of what seems at first blush a very limiting approach. There are several papers on simulation in game theory, often embedded in a CA framework. An example is provide by Kirchkamp's paper on the spatial evolution of the Prisoner's Dilemma (Spatial Evolution of Automata in the Prisoner's Dilemma, pp. 307-358).

Another major field in micro-simulation (Distributed Artificial Intelligence) also merits a section in the book, with contributions from Doran (Simulating Societies Using Distributed Artificial Intelligence, pp. 381-393), Conte and Castelfranchi (Simulating Multi-Agent Interdepencies: A Two-Way Approach to the Micro-Macro Link, pp. 394-415) and Manhart (Artificial Intelligence Modelling: Data Driven and Theory Driven Approaches, pp. 416-431). Repeating what seems to be a pattern in the book, Doran's paper provides a considered overview of the value of D.A.I. for social simulation.

As an empirical sociologist, sympathetic to approaches based on methodological individualism, I find this volume quite encouraging. It shows a good level of social scientific sophistication among simulators, who in turn demonstrate that the technologies they possess and develop are more and more useful for real social science problems. I am pleased to see elements of a well-informed discussion emerging on the hoary old macro-micro (structure-agency) problem in sociology. Conte and Castelfranchi even promote it to the status of a three-letter acronym (MML for macro-micro link) while criticising Giddens who neither invented nor solved this problem by labelling it "structuration". In particular, it is good to see an argument that we must be neither (exclusively) bottom-up or top-down in this regard. It is too easy to get infatuated with emergence and forget that the fascination should be in the recursion.

The volume is a collection of conference papers and suffers some of the inevitable problems this implies: some deviation from the theme, some variability in quality, sometimes less polish than other media might demand; but conference proceedings have advantages too, especially when the conferences happen at a time when the discipline is developing productively, when they involve a lot of good researchers and when there seems to be effective communication between the participants. This is a very useful volume to have around at the moment for anyone interested in simulation of social phenomena.


ButtonReturn to Contents of this issue

© Copyright Journal of Artificial Societies and Social Simulation, 1998