Reviewed by
Michael Mäs
ETH Zürich
In the past decades, the social sciences experienced an explosion of simulation modelling. However, modellers apply very different strategies when they develop, implement, analyse, and report about simulation models. In fact, despite the success of computer simulation there is still no consensus on very basic issues, such as the appropriate complexity of simulation models. Should they be as simple as possible, or should we exploit the ability of computers to represent even very complex systems? In addition, simulation modelling often faces scepticism from scholars who question the usefulness of formal modelling in general and also by researchers who focus on analytical tools to analyse formal models. The handbook "Simulating Social Complexity" identifies, compares, and debates existing approaches to simulation. It consists of 26 chapters that are organized in 4 parts (Introduction, Methodology, Mechanisms, Applications). Here, I point to the highlights of the methodology and the mechanisms parts, which are the essential parts of this handbook.
Chapter 4 by Emma Norling and colleagues provides a very good introduction to the informal approach to simulation modelling, which is the approach used by most modellers. Importantly, this chapter identifies typical challenges and sources of errors and proposes several general rules that can help prevent problems. With more detail, chapter 6 by José M. Galan and colleagues summarizes possible types of errors and artefacts that can occur during the modelling process. In addition, the authors discuss methods that help avoiding and detecting errors and artefacts. In chapter 7, Volker Grimm and colleagues describe the ODD protocol, a standardized method for the documentation of simulation models. The authors illustrate ODD with the Bounded Confidence model. Chapter 8 by Nuno David provides an overview over methods of validating and verifying simulation models. Importantly, the author guides the reader in developing models that can be validated.
In Chapter 9, Andrew Evans and colleagues show various graphical, statistical, and index-based techniques to quantify and illustrate the phenomena and dynamics that simulation models generate. The authors discuss advantages and disadvantages of different methods and point to available software.
My personal highlight of the handbook is Chapter 11. In this chapter, Luis R. Izquierdo and colleagues analyze the relationship between mathematical methods and computer simulation and show how both methods can benefit from each other. For instance, one can gain a better understanding of simulation models by studying parts of the model with analytical tools. Likewise, simulation tools allow the modeller to relax assumptions of analytical models even when this implies that the model is no longer analytically tractable. This can be very helpful when studying whether a given ingredient of an analytical model is responsible for a model prediction. The authors argue that simulations should be developed in a way that the model or at least parts of the model can be translated directly into an analytical model. This is shown in a very intuitive example with a Markov chain model.
Part III of the handbook is an essential part, as it reviews different theories of human behaviour and models of human populations that have been integrated in simulation models. In particular, modellers who do not have a background in sociology, psychology or economics will find this part very helpful.
In Chapter 13 Guido Fioretti provides a very good and comprehensive overview over the theory of rational action and game theory. The author addresses the core assumptions, applications and implications of this general approach to individual behaviour and discusses its advantages and weaknesses. Chapter 15 by Francesca Giardini and colleagues focuses on reputation mechanisms and their role in models of the emergence of cooperation and partner selection in market settings. This is a good introduction to the various approaches, theories and domains of this important and growing field of simulation research.
In Chapter 16, Frédéric Amblard and Walter Quattrociocchi review techniques of modelling networks and geographical space. This chapter presents alternative ways to model and analyse geographical space and social networks. It also covers typical dynamics that have been studied with these models, such as the small world effect.
Chapter 17 discusses another central theory of individual behaviour, learning theory. Michael Macy and colleagues summarise and discuss classical and contemporary learning models and show how learning has been implemented in simulation models. This is a very good chapter. It also discusses the differences between learning theory and alternative theories of individual behaviour, such as the theory of rational action. In Chapter 18, Edmund Chattoe-Brown and Bruce Edmonds give a very good introduction to the key concepts and assumptions of evolutionary theories. They show and discuss how these biological theories can be translated into sociologically meaningful models.
Considering its high price, I do not expect that many readers will add this handbook to their personal library. This is a pity, as it consists of many very good chapters and can, thus, be recommended to anybody who wants to learn more about good practice in simulation modeling.
Return to Contents
of this issue