Browse Books

Reviewer: David H. Withers

This fascinating book touches lightly on a diverse collection of computer simulation models. Casti also presents his rules for good models. It is hard to characterize the audience for this book. It is not a classroom text, a reference manual, or a research paper. It would provide background material for a simulationist looking for an easy-to-read summary of the diversity of advanced simulation applications. But it is likely that simulationists will not find many familiar topics. The author's affiliation with a particular research institute is obvious, and many of the examples come from his personal knowledge of work at that institute. Most of the models are introduced by a discussion of the problems in the particular area of study. The discussions are heavily laced with the author's viewpoints and philosophies. Each introduction is followed by a high-level description of the model. No mathematical or technical background is required to understand the problems and models. The models cover a diverse set of areas, ranging from traffic congestion to DNA folding. Each chapter ends with a comparison of the models to the definitions of a “good model.” Most of the models rate fairly high—of course, they were selected by the author for discussion. Chapter 1 describes models used to simulate professional football, the stability of planetary systems, and economic systems. Casti describes the different roles that models play: predictive, explanatory, and prescriptive. Brief examples of each are provided. He defines good models as ones that are simple, clear, free of bias, and tractable (affordable to compute). In chapter 2, “Pictures as Programs,” Casti discusses how seemingly complex two-dimensional objects can be generated by fairly simple algorithms. Examples include biomorphs and artificial plants. A random collection of short discussions on Turing machines, ASCII code, dialogues created via artificial intelligence, and electronic computation follows. These examples lead to descriptions of using protein folding for computing and of the difficulties in building a global weather model. Casti makes the point that the initial use of computers was for high-speed computation of complex mathematical models, but the emerging uses include symbolic computation for language, music, and art. The chapter closes with examples of models of political strategies and the stock market. The main point seems to be that we can now investigate systems where real-life experimentation is not economically or politically possible. The third chapter focuses on models where there is an apparently paradoxical difference between the expected and the actual results. Casti defines chaotic machines as those in which small changes in inputs produce dramatically different behaviors. The examples are selected to support his definition, but are flawed. In each case, the reason for the apparently unpredictable behavior is a lack of detail, precision, or both in the model. In other words, the models are not sufficient for the purpose. Readers who are unfamiliar with modeling dynamic systems may draw erroneous conclusions about complex systems, believing that they must display surprising behavior, and will come away with the wrong impression of our current ability to model dynamic systems. This chapter continues with a long discussion of computing machines, including a DNA machine. There is also a section on paradoxes, which is interesting but not relevant to the central theme. Examples presented in chapter 4, grandiosely titled “Artificial Worlds,” include simulation models of city vehicle traffic and war games. The war games models are interesting, but do not represent the state of the art in military and political strategy analysis. They do represent a unique kind of model—those which include some intelligence so the actors can learn from their environment. Casti maps these digital representations to biological evolution. The last chapter summarizes the previous chapters and presents a mapping of mathematical and computational models to the real world. The discussion is interesting. The book also includes a modest bibliography. This easy-to-read book provides some cultural background for simulationists, and may encourage readers to think beyond their current paradigms. Its major defect is its treatment of complex and chaotic systems, which may mislead some readers.