Reviewed by
Martin Neumann
Bayreuth University
The book presents a collection of papers based on talks accepted for the International Workshop on Multi-Agent-Based Simulation, in Estoril, Portugal held on 12 - 16 May 2008. Due to the fact that the book is based on a meeting of researchers in multi-agent based simulation, it is geared towards practitioners rather than being an attempt to convince a general academic audience of the benefits of agent-based simulation. Despite the existence of periodicals such as the Journal for Artificial Societies and Social Simulation or Computational and Mathematical Organization Theory, this book series still provides a valuable archive for the documentation of the constant progress made in social simulation. The present book is a snapshot of the current state of the art, with contributions from Europe, Asia and America. Beside models of economic and social behaviour, a number of contributions address methodological and technological issues. The book focuses on methodology and technology. Particular striking is the interdisciplinary account discernible in a number of papers as well as in the selection of the diverse topics. On the other hand, this interdisciplinary approach means that the paper are rather heterogeneous. The anthology is organised in five sections: the first section consists of two papers dealing with the "simulation of economic behaviour". The second section covers "modelling and simulation of social behaviour", followed by sections on "applications" and "techniques, infrastructure and technologies". The final section explores "methods and methodologies".
The section on economic modelling includes a paper that describes a model of a theory which incorporates cultural aspects into economic behaviour and an evidence-based model, rejecting classical theoretical accounts with qualitative evidence from the financial market.
The section on simulation of social behaviour consists of three papers: a model of the spread of HIV, based on empirical evidence, an abstract model of opinion dynamics, and a model presenting an architecture and first implementations of a model of an organisational network structure with cognitively rich agents. It is applied to asymmetric warfare and validated against data from terrorist groups.
The section on applications comprises three rather heterogeneous papers. One paper explores cluster analytical methods to evaluate data of student responses to multiple choice tests. Data partitioning is used to train software agents that exhibit behaviour typical for each cluster. Another paper uses simulation technology to reduce operational costs for container terminals. A simulation of the optimisation of how to unload ships is presented. In a further paper malware, such as viruses, worms or Trojan horses, are simulated. Agent-based models are made to obtain a more realistic representation of computer networks to study the proliferation of infectious software.
The section on techniques, infrastructure and technologies ranges from a report on the design of a participatory study of environmental management to an abstract semantics of event-based multi-agent simulations and a position paper suggesting a bridge between the micro macro gap by an inclusion of norms, personality and culture in a highly sophisticated cognitive design of agents. The aim of this approach is to furnish support for the evaluation of policy measures. Macro models do not suffice, it is argued, because the effect of policy measures depends directly on individual behaviour.
The final section on methods and methodologies consists of four papers. Firstly, a quantitative method for comparing simulation results is introduced. By including cluster analytical methods, the paper points in the same direction as the first paper in the section on applications. Moreover, a data-driven simulation of the change of moral values in Spain and a comparison of views from physics and agent-based modelling on complexity are reported. The failure to replicate an equation-based model into an agent-based model is documented. Finally, a post-modern approach to simulate the dialectic between emergence and social causation is elaborated. By building on embodied cognition, this work concentrates on communication and symbolic interaction to simulate the generation of meaning through sense-making.
Given the heterogeneity of the papers, the reader's background may determine what appears to be of most interest. Given the sample of contributions to the workshop, a certain degree of arbitrariness in the sections is inevitable. However, a trend towards enriching the cognitive complexity of the agents seems discernible. This is in line with an observation that can already be made in the comparison of the two papers in the economic section: they highlight very divergent applications of agent-based models. On the one hand, simulation allows us to overcome mono-causal theories by including theoretical insights from other domains in the common framework of the simulation model. On the other hand, modelling provides a tool to encrypt data gathered with ethnomethodological studies without any intervening theory. Simulation thus provides a framework to study implications of ethnomethodological field work. Both accounts, however, diverge from the KISS principle. The same observation holds for the cognitive complexity of the agents' architectures presented in a number of contributions to this volume. This, however, calls for a comparative analysis of the whole book series to identify whether there is a trend in agent-based research towards abstaining from the KISS principle.
Return to Contents
of this issue
© Copyright Journal of Artificial Societies and Social Simulation, 2009