Reviewed by
John Bragin
UCLA Human Complex Systems Program
This is a useful volume. Its usefulness is in the number and variety of its chapters, but, despite the wishful thinking of the title, there is no synoptic vision that unifies these 77 chapters categorized in three parts ("Methods", "Models", and "Applications",) nor is it clear why many of the chapters are placed in one of these parts rather than another, or why just these three parts alone have been chosen.I begin this review with a brief criticism of the claims for unification inherent in the title and then discuss some of the insightful and useful contents of the collection. (Only "some", because with 77 highly diverse chapters, to do more is not possible in principle or in practice. Most of the chapters I highlight deal with urban issues, a particular interest of mine.) There is a short introduction to the volume, but it is general and does not give any overview of the structure or contents of the volume - as most introductions to scientific collections usually do.
A sentence from the "Introduction" claims: "Our science can provide a unified foundation and framework for the study of all complex systems." This is a bold statement, not only because there is as yet no Kuhnian paradigm for the sciences of complexity (and there may never be one), but also because the fragmentary nature of this collection belies the claim. There does exist a loose and useful basket of concepts and methods across the sciences of complexity (such as complexity, emergence, evolution, adaptation, and self-organization listed in the "Introduction"), but there is no widely agreed-upon "unified foundation and framework" that defines (not to say quantifies) any of these. Page xi lists five "unifying themes in complex systems" which includes those from the "Introduction", plus informatics and dynamics. Although some aspects of several of these concepts (such as evolution by natural selection, much computer science as it relates to informatics, and the mathematics of several types of dynamical systems) are well defined both qualitatively and quantitatively, most of what belongs to complex systems science per se is still ill-defined (or multiply-defined with no congruence among the definitions). And there exists a number of methods (such a network modeling, neural nets, stochastic analysis, game theory, time series analysis, and agent-based simulation) that are by no means unified or even satisfactorily integrated with each other. Complex systems science is not only in its infancy (as the "Introduction" does point out), but it is also in a disunified, pre-paradigm stage, with a wealth of ideas, methods and investigations that are well-represented in this volume.
Each of the chapters is a summary of one of the presentations given from June 25 through June 30, 2006 at the Sixth International Conference on Complex Systems, but not all the presentations (listed on pages xii though xxiii) are represented. The chapters are short (six to eight pages) and highly condensed, which does not necessarily mean they are inadequate. Each of the days was divided into a number of topic or domain-specific sessions, some of which were repeated on subsequent days with different approaches by different speakers. These included issues in modeling, networks, homeland security, mathematical methods, engineered systems, innovation, biology (cellular, ecological and evolutionary), social systems, education and health care, physical systems, game theory, and global systems. The volume itself is not structured along the topics or domain-specific sessions of the conference, but it would have been more perspicuous had the editors done so, rather than the three "Parts" into which the collection has been divided.
Part I: Methods (25 chapters) begins immediately with a chapter on an information-theoretic characterization of emergence by Daniel Polani. (None of the three parts of the volume has an introductory overview by the editors of the structure or content of the part, which is standard practice for scientific collections. But, since none of the parts really has any uniformity of structure or content, this is not surprising.) This chapter is followed by three others that deal with different notions of emergent complexity. All of these chapters contribute to the on-going discussion of just what emergence and irreducibility amount to. Next come several chapters dealing with modeling, the most interesting of which is Chapter 6 by Tibor Bosse, Alexei Sharpanskykh and Jan Treur, which describes the LEADSTO language for combining agent-based and dynamical systems approaches. Then come a number of chapters dealing with theoretical and empirical aspects of network science, including questions of innovation, economic geography, and terrorism. Chapter 23 ("Complexities, Catastrophes and Cities" by Giuseppe Narzisi et al.) describes an agent-based model for urban disaster management (natural and human-made). Cities cannot be experimented on in laboratories. No linear, equilibrium, math-based models (seeking reliable predictive accuracy) are going to be of much help in the planning, preparation and implementation of ways to meet disaster or emergency situations in the metropolis. Thus continued work on agent-based simulations seems to be the best and brightest hope to meet these problems.
Part II: Models (31 chapters) contains pieces on biology (from cell systems to ecological systems, including cooperation and conflict); pedestrian crowds, network [WWW, etal] traffic models; systems engineering; industry, markets, global economics & sustainable development; and game theory (Nash equilbria). Chapter 14 of this part ("Simulation of Pedestrian Agent Crowds, with Crisis" by Margaret Lyell, Rob Flo and Mateo Mejia-Tellez) presents a model of agent behavior in crowds where the agents are faced with a crisis, such as a fire, that may impede their goals. Pedestrians have cognitive (including belief-structure), affective and locomotor capacities. Police officer agents act on the basis of rules and training. The goal of the authors is the creation of a user-friendly simulation framework so that "the user/analyst will be able to 1) develop the geometry of the urban area using a template; 2) develop realistic agent personalities using templates; 3) assign resources to the scenario, including police officer agents; and 4) construct variations of the scenario, for different simulation investigations."
Part III: Applications is a potpourri of chapters dealing with such domains as sensor networks, engineered systems, business management, language, international affairs, health and vehicular traffic. Chapter 2 of this part ("An Exploration into the Uses of Agent-Based Modeling to Improve Quality of Healthcare" by Ashok Kay Kanagarajah et al.) is a contribution to the discussion of health care delivery as a complex adaptive system. Health care is a relatively new area in which complexity concepts and methods are being applied. Given both the necessity of delivering wide-spread and effective care and the growing costs of such care, complexity science (and particularly agent-based modeling) may be our best approach to meet this need and to control costs as best we can, with as little compromise as possible in the quantity and quality of the healthcare we can provide. Another pressing problem for the societies of large urban areas is traffic management. Chapter 6 of this part ("Self-Learning Intelligent Agents for Dynamic Traffic Routing on Transportation Networks" by Adel Sadek and Nagi Basha) presents a Reinforcement Learning approach to Dynamic Traffic Routing, in which an agent that will route traffic is trained first in a simulation model and then deployed in the real world where it refines and updates its control policies over time. The authors used a Cell Transmission Model, where the system is divided into discrete cells across which a vehicle moves. [This discrete model is simpler to implement than a continuous traffic system, but it might also fail to be complex enough to be ultimately useful for real-world traffic handling]. The authors propose further research using neural networks that "will allow us to deal with a bigger set of states as well as achieving a smoother and continuous representation of the state space".
Because this volume offers such a wide range of concepts, methods, and applications, anyone involved in the sciences of complexity will find many of these chapters stimulating. I have chosen a handful of chapters to highlight that deal with urban issues, which is one of my main areas of complexity interest. Others will find other topics and domains interesting and useful in their own and allied fields.
Return to Contents
of this issue
© Copyright Journal of Artificial Societies and Social Simulation, 2010