© Copyright JASSS

JASSS logo ------

Social Self-Organization: Agent-Based Simulations and Experiments to Study Emergent Social Behavior (Understanding Complex Systems)

Helbing, Dirk (ed.)
Springer-Verlag: Berlin, 2012
ISBN 9783642240034 (hb)

Order this book

Reviewed by John Bragin
UCLA Lecturer (periodic) in Complex Systems Science

Cover of book Agent-Based Modeling is the core method of complex systems science. It embodies Joshua Epstein’s generative social science motto: "If you didn’t grow it, you didn’t explain its emergence". This volume is a collection of models whose aim is not only to better explain a variety of phenomena in the social, cultural and economic realms using agent-based models, but also to manage these complex phenomena. In the "Preface" Dirk Helbing says of the models in the book: "it seems that many characteristic features of complex social systems can be understood from simple models of social interactions". Most of the models also involve evolutionary game theory.

The book falls roughly into three parts, although the editor has not formally divided the volume into sections. Chapters 1 and 2 introduce modelling in general for socio-economic systems and agent-based modelling in particular. Chapters 3 through 13 deal with specific social situations. And chapters 13 through 16 discuss the systemic risks in social complexity and ways these might be managed. Excellent, ample bibliographies follow each chapter.

Chapter 1, “Modeling of Socio-Economic Systems”, focuses on the constraints and opportunities of formal models and simulations. It is short, as it must be as an introduction, but it is an excellent piece. Helbing lists 15 particular difficulties; recommends and gives guidelines for beginning with a qualitative, narrative descriptive-model; goes on to describe simple models and modelling complex systems. He then discusses the challenges of socio-economic modelling, including those of computational experimentation; the limitations of models; and the different interpretations of models. He ends with thoughts on the need for probabilistic models and more than one model for each target system; and for the collaboration between natural and social scientists.

Chapter 2, “Agent-Based Modeling”, is now a regular part of the reading list for my introductory courses in complex systems science. There are many other papers on ABMs, but few have the materials covered in several sections of this chapter. For example, Section 2.3, “Implementation and Computer Simulation of Agent-Based Models”, includes a subsection “Common Mistakes and Error Sources”. Helbing begins this chapter with an answer to the question of why one should use agent-based models, especially in contrast with equation-based models; he then discuss the principles of ABMs; their implementation as a computer simulation; their application, potential and limitations; and ends with some thoughts on the future of the method.

Chapters 3 through 13, include models on pedestrian crowds, opinion formation, cooperation, the evolution of morality, coordination and competition, conflict, and decision responses to information. A number of chapters deal with the use of agent-based models to tackle spatial situations, a topic where ABMs are particularly strong in comparison with macroscopic equation-based models such as those used in systems dynamics models.

Given an initial homogenous distribution of agents and a game theoretical model applied to neighbours, what is the affect on the dynamics of the spatial self-organization of success-driven mobility, depending on different payoff matrices and noise levels? This is the question of Chapter 5. Chapter 7 shows how the “Prisoner’s Dilemma” game can be used to show a “Co-Evolution of Social Behavior and Spatial Organization”. In 1992 Nowak and May published a seminal paper, “Evolutionary games and spatial chaos”, a computational study that showed that altruistic, cooperative behaviour would persist through social clustering. In Chapter 7, “Social Experiments and Computing”, Helbing and Wenjin Yu review a recent computational study of a spatial prisoner’s dilemma game (building on the Nowak and May paper) in order to discuss the various ways social experiments should and could be done in the future, using the help of computers.

Chapter 14, “Systemic Risks in Society and Economics”, might also be included in the third part of the volume (as I informally divide the book), along with Chapter 15, “Managing Complexity”, and Chapter 16, “Challenges in Economics”, because although all three chapters do not deal exclusively with Economics (including Finance) the problem of Economics as an Evolving Complex System is central to the ideas presented in all three. We are now faced with a globe-spanning economic-financial crisis that may be in its length (if not its depth) ruinous in its effects greater than that of the Great Depression.

There can be no doubt that the hegemony of the Neo-Classical Theoretical stance in academia, government and business is in great part responsible for our current condition. In Chapter 14, Helbing addresses various alternatives to this theory including power laws; network interactions and failure cascades; self-organized criticality; non-linear systems patterns; obstacles to traditional methods of control; financial market instability; lists some ways to manage complexity; and reducing network vulnerability. Helbing gives many of the commonly discussed aspects of these, but also provides additional insights of his own.

“Managing Complexity”, Chapter 15, by Helbing and his colleague S. Lammer is notable for its focus on “Some Common Mistakes in the Management of Complex Systems” and the ways in which self-organization can be guided rather than systems controlled.

Finally, Chapter 16, “Challenges in Economics” by Helbing and Balietti, written in 2010, begins with a list of ten real-world challenges any approach to economics (including finance) must meet. The authors hen address the following problems with traditional economic theory: Homo Economicus, the Efficient Market Hypothesis, the Equilibrium Paradigm, the Prevalence of Linear Models, the Representative Agent Approach, the Lack of Micro-Macro Link and Ecological Systems Thinking, the faulty use of Optimization, the classical Control Approach, the failure to include such psychological and social Human Factors as emotion and norms, and finally the failure to consider information as an explanatory variable. The chapter ends with some consideration of the role of other scientific fields in economic thought: econophysics, ecology, computer science, and the other social sciences.

The chapters need not be read in any particular order, however I recommend that chapter 1 and chapter 2 be read as a unit; that chapters 5 and 6 be read before chapter 7 - as chapter 5 deals with mobility in geographical space and chapter 6 with social cooperation and chapter 7 combines the two; that chapter 12 be read before chapter 13 - as they both deal with learning; and that chapters 14, 15 and 16 be read as a unit.

All of the chapters in the book were previously published as papers or book chapters that were authored or co-authored by Dirk Helbing. They date from 2000 to 2012, with most of the chapters from the last few years, several from around 2004-05, and one from 2000.

In the “Preface” Helbing does not give a reason for the choice of the chapters he includes, beyond the fact that they are meant to show how simple, agent-based models of social systems (often using an evolutionary game-theoretical framework) are yet capable of capturing the recognizable complexities of the real world. Chapters 1 and 2 (introducing modelling in general and agent-based modelling specifically) form a natural introduction which need no justification Chapters 14, 15 and 16 (dealing with socio-economic risks in general, managing complexity, and challenges in economics) are all of an extensive nature and concern problems for which agent-based modelling is considered by us complexians the best available modelling tool. But why just those topics and those chapters - and not others - were chosen for the rest of the volume, we do not know. Of course, many of the topics, such as pedestrian dynamics, opinion formation, and cooperation are prominent topics of complexity research. But some other topics, such as Chapter 5 on success-driven mobility, and some of the treatments of learning in Chapters 12 and 13 are a bit recondite, at least to this reviewer. At any rate, I would have liked for Dirk Helbing to have given us more on why what was included was included, what he had to leave out and why, and what he would have liked to include, but couldn’t.

That said, I recommend this as an important addition to the growing work on both the theoretical and practical world of agent-based simulation modelling in the social sciences.


ButtonReturn to Contents of this issue