Reviewed by
Pietro Terna
Dipartimento
di Scienze economiche e finanziarie, Università di Torino,
Italy.
Like James March, who declares in its foreword "... I am hardly an unbiased commentator on the present book", I am hardly an unbiased reviewer. Having worked for many years to develop and diffuse simulation models in economics, I read Lomi and Larsen's book with a great pleasure, discovering new view points about both organisations and simulation techniques.
Firstly, the foreword by March provides many helpful insights that explain the difficult journey which simulation has made to become a standard tool in organisation science (and in social science too).
March argues that by the mid 1960's, a plausible expectation would have been that computer simulation would either relatively quickly be accepted as part of the canon of social science methods or be discarded equally rapidly as a methodological and theoretical dead-end. In fact, neither of these outcomes occurred.
In fact, simulation was placed in limbo and its development has been in some sense protected by historical events. This provides interesting evidence for the importance of "protection in evolution". March recalls that one of the grander dilemmas of adaptation is that novel developments require a social structure that buffers them from short-run and local selection pressure so they can survive. In the case of simulation, three features of the social organisation of disciplinary knowledge and simulation technology led to this kind of (spontaneous and unintended) protection.
First, there was no traditional social science discipline in which simulation was a serious contender for disciplinary pre-eminence (...) In that sense, computer simulation survived in the social sciences by being almost invisible and largely irrelevant.
Second, simulation modelling endured particularly in places shielded from disciplinary orthodoxy. (...) simulation modelling (...) maintained itself by defining itself as outside the ken of the major disciplines.
Third, computer simulation attached itself to the larger community surrounding computers. The large budgets, extraordinary hype, and palpable importance of computing and information technology made association with those fields beneficial. (p. xiv)
This description of niche survival is quite realistic and (in this reviewer's opinion) makes it a matter of urgency to increase the rate of growth in simulation applications. Also as a reviewer's remark, I remind the reader of the strong internal critique presented by Pryor (2000). In his ironic paper, an unknown author from 2028 (when an asteroid has crashed on the Earth), looks backwards and observes "that in a typical complexity book in the late 1990s (. . .) almost all of the essays have no real empirical applications, aside from a few interesting anecdotes."
Lomi and Larsen's rich introduction to the book shows that results are coming. Recalling the arguments of Coleman and Fararo (1992), Lomi and Larsen consider that no completely satisfactory theory of organisation exists because no theory is available that simultaneously (a) explains the behaviour of organisations, rather than the behaviour of selected individuals (or groups) within organisations; (b) provides a model of individual choice and motivation, and (c) establishes a clear translation mechanism between the level of individual behaviour and that of organisational behaviour (and back again).
The book avoids the temptation to classify restrictively, as an a priori choice, the missing explanations in terms of complexity, self-organisation and emergence. It introduces instead a "lighter" classification in terms of "rediscovering problems", "framing arguments" and "taking views".
Many different applications of simulation to organisation theory (summarised in Table 1) can be found in this collection. I have also used this table to present a personal judgement on the difficulty of replicating the simulation results. Obviously, one could ask all the authors to provide the code they used in doing their simulations, but this simple request can create a lot of problems. My experience was that when I did not attach the code I used to my publications (along with the instruction to run it), I had enormous problems in replicating even my own results from only a few years before. The programming environment is not the same, some files are lost, many conditions have been changed in recovery from computer crashes and so on.
So, it would have been better if the editors of such an important book had obtained the code used in the papers from the authors. I know, however, having been an editor, that this is not an easy task!
A collection of papers, even if well organised and thought out, is always difficult to conclude. This book brilliantly avoids that difficulty thanks to an important "Afterword" by Richard M. Burton, providing a deep analysis of three key characteristics of simulation: specification, versatility and efficiency. A reviewer's synthesis follows:
Table 1. The Contents of the Volume
| Author/s | Content | Methodology | Difficulty of Replication |
| J. R. Harrison and G. R. Carroll | Culture in Organisations | Computational model | Medium/High |
| K. M. Carley and V. Hill | Structural Change and Learning | Agent based with an identified model | Medium/Low |
| M. W. Macy and D. Strang | Innovation and Followers | Computational model | High |
| C. H. Loch, B. A. Huberman and S. Ülkü | Status Competition and Group Performance | Computational model | High |
| M. Prietula | Advice, Trust and Gossip | Agent based | Low |
| M. S. Bothner and H. C. White | Market and Monopoly | Computational model | High |
| D. N. Barron | Dynamics of Organisational Populations | Computational model with agent flavour | Medum/High |
| D. Krackhardt | Diffusion of Controversial Innovations | Computational model | High |
| A. Lomi and E. R. Larsen | Age Dependence in Failures | Cellular automata | Medium |
| J. H. Miller | Evolving Information Processes | Graphs | High/Medium |
| D .A. Levinthal | Adaptation on Rugged Landscapes | Fitness landscapes with agent flavour | High |
| F. Malerba, R. Nelson, L. Orsenigo and S. Winter | Product Diversification | Computational model | High |
| M. A. Sastry | Dynamic Complexity in Organisational Evolution | System dynamics | High |
| L.Pólos and M. T. Hannan | Nonmonotonicity in Theory Building with Applications | Qualitative formal analysis | Not applicable |
Specification: Simulation requires us to specify the world we want to investigate. It can be rich and complex or it can be simple. It can begin from simplicity and develop complexity. We must specify the "black box"; we cannot just assume it exists. In simulation, we provide behavioural specifications rather than making behavioural assumptions and thus reduce the gap between models and theories. The central issue is that we know more about the rich simulated world than is usually the case when we use the "real" world as our laboratory. Within this necessary specification, the simulated world is a laboratory in which we know important parameters because we specified them; we did not merely assume them.
Versatility: The rich world of simulation is versatile; we can perform many different kinds of studies: test hypotheses, explore new ideas, create large amounts of data, help solve problems and go outside the boundaries of the "real" world. The rich simulated world can thus be used to understand the limits of our "real" world; it can be extended to investigate the limits of the possible; and it can give a picture of the likely, of what might be.
Efficiency: Simulation is efficient, or relatively easy to use for some types of studies and questions. This is especially true for systems that are complex, with a large number of variables and interdependencies; or dependent as a result of path sequences, evolution or historical events. When compared to "real world" observation and experimentation for complex, evolutionary systems with feedback, simulation is inexpensive. It is a versatile laboratory in which we can specify relations that are complex, path dependent and involve feedback; we can also calibrate experiments aimed at generating different new and plausible worlds, and explore the results in terms of their effects on organisation.
To conclude, this book is highly recommended both for simulation theorists and for practitioners, even if they are not directly engaged in the organisational field..
ReferencesCOLEMAN J. and T. Fararo 1992. Introduction. In J. Coleman and T. Fararo, editors, Rational Choice Theory: Advocacy and Critique. Sage Publications, Newbury Park, CA.
PRYOR F. L. 2000. Looking backwards: Complexity theory in 2028. In D. Collander, editor, The Complexity Vision and the Teaching of Economics. Edward Elgar, Cheltenham.
Return to Contents
of this issue
© Copyright Journal of Artificial Societies and Social Simulation, 2002