Order this book
UCLA Lecturer (periodic) in Complex Systems Science
The authors then discuss some particular problems of modeling CAS and identify and answer four questions "fom the less abstract to the more abstract". Each of these topics could occupy a long research paper in itself, but the authors offer some interesting insights in a paragraph or two for each question: (1) How to better develop models of the interaction data from various components of CAS?; (2) How to describe CAS to facilitate communication across scientific and disciplinary boundaries?; (3) In the case of a dearth of real-world data, how can CAS simulation models be validated primarily using meta-data or concepts?; (4) In general, how can multidisciplinary CAS research projects be structured and executed based on an availability of resources and commitment?
The chapter ends with an explanation of the motivation for the series, namely fulfilling the need for a unified framework, the aims and objectives of the series as a set of guidelines and an overview of the volumes that present the guidelines coupled with case studies.
Chapter 2, "A Unified Framework", first presents an overview of the proposed framework. The framework is then described from two points-of-view: First from the point-of-view of the objective and level of commitment of a CAS case study, and second from a point-of-view that concerns the types of data available for such a study. In terms of the overview, there are four levels to the framework: First a level of network modeling. Second, a level of agent-based modeling. Third, a new, non-textual and quantitative Descriptive Agent-Based Model approach (DREAM) "using a combination of pseudo-code based specification, complex network model and a quantitative model 'fingerprint' based on centrality measures of the agent-based model which are all associated closely with the ABM" and which the authors propose as an alternative to such text-based descriptions as the ODD (Overview, Design Concepts and Details) protocol. And fourthly, an in-simulation verification and validation method for agent-based models, which the authors call the Virtual Overlay Multiagent Systems (VOMAS) software engineering, approach. All of these to be described in detail in the forthcoming volumes. Although it is evident why the first of these two are levels (CNs are more general than ABMs), I do not understand why the second and third are levels, or how they are related to the first two, except that the third is a description and the fourth a verification/validation. Perhaps this is made clear in the following volumes.
In addition, it is not evident to me that these four "levels" unify the framework or that they unify the field of CAS modeling. It still appears to me that they are presented as four complementary methods: The first two are modeling methods per se , the third a method for description (not modeling) of CAS) and the fourth verification/validation (again not modeling) of models of CAS. This first volume contains valuable insights, but the ultimate value (usefulness) of the series will be in the details of the individual volumes that deal with each of the four "levels" (which I prefer to call methods) of the framework.
In two flow charts the authors then present the decision-making processes for choosing among the above four framework levels in relation to CAS research study objectives and in relation to available data types. It seems that the authors imply that there are some situations in which complex networks alone or agent-based models alone will be the appropriate methodology for modeling a CAS. However, I think this is a flawed approach. Although the two methods have developed mostly in isolation for some time and significant strides have been made, it seems to me that further progress requires they now be united. All social agents are embedded in networks and all networks face the problem of nodes as black boxes unless the nodes as autonomous proactive agents are given at least elementary perceptive, learning, cognitive-emotional and behavioral traits. So, if we are to move forward in using ABMs and CNs to address problems in the real world, the two methods need to be joined for even the simplest real-world situations. In addition, other methods such as genetic algorithms, neural networks and evolutionary game theoretic methods need to be drawn on, but it appears that the authors do not consider these.
Chapter 2 ends with an outline of the upcoming volumes: These volumes will be co-authored by Niazi and Hussain and are scheduled to appear as follows: Volume II: Cognitive Agent-based Computing-II: Modeling, Visualizing and Analyzing Complex Adaptive Systems Using Complex Networks, expected publication date: Dec 1, 2013; Volume III: Cognitive Agent-based Computing-III: Exploratory Agent-based Modeling for Complex Adaptive Systems, expected publication date July 1, 2014; Volume IV: Cognitive Agent-based Computing-IV: Descriptive Agent-based Modeling for Complex Adaptive Systems, expected publication date Dec 1, 2014; and Volume V: Cognitive Agent-based Computing-V: Validated Agent-based Modeling for Complex Adaptive Systems, expected publication date July 1, 2015. (The dates were given to me in an email from Muaz Niazi.)
In Chapter 3, "Complex Adaptive Systems", the authors give their view of CAS and their view of the key characteristics of CAS (from John Holland), and some specific examples of CAS. Their two exemplars of CAS are life and human social systems. The seven basics of CAS drawn from Holland’s 1996 book Hidden Order: How Adaptation Builds Complexity are divided into properties and mechanisms: "Aggregation" (property), "Tagging" (mechanism), "Non-Linearity" (property), "Flows" (property), "Diversity" (property), "Internal Modeling" (mechanism), and "Building Blocks" (mechanism). In addition to these properties and mechanisms, CAS are characterized by emergence, which the authors follow Yaneer Bar-Yam in defining: "Emergence refers to the existence or formation of collective behaviors - what parts of a system do together that they would not do alone". Four examples of CAS are then briefly described: 1) Plant growth; 2) Scientific communication (principally journal publications, supported by conferences and meetings); 3) Complex Adaptive Communication Networks (the use of the Internet and WWW for social, financial, corporate, gaming and other purposes); and 4) Simulations of Bird Flocking Behavior.
Chapter 4, “Modelling CAS”, focuses on the concept of modeling in general, on agent-based modeling in particular, on Netlogo, the verification and validation of ABMs, the theoretical basis for complex network modeling, some quantitative measures involved in network modeling, and lastly on references to the literature on software tools for network modeling. Given the limited amount of space available in this introductory volume of the series, the authors do a nice job here in their brisk overviews. All these topics are to be covered in more depth in the upcoming volumes.
At a list price of $49.95 this 50-page, paperback volume is far too expensive. It is not a book that one would buy and keep for future reference. It is well worth careful reading, but hardly worth adding to ones own library. I recommend that, if you are at a college or university, you get your librarian to purchase it and the up-coming four volumes. Not having yet seen any of the up-coming volumes, I cannot say if any of these would be worth purchasing and keeping for long-term use.
As part of the SpringerBriefs series, individual volumes must be between 50 and 125 pages long. Each of the up-coming volumes will be from 70 to 80 pages long and will contain more detail, in the form of worked case studies, to support the Unified Framework Guidelines. (This information in an email to me from Muaz Niazi.)
Return to Contents of this issue
© Copyright JASSS, 2013