Browse Books

Reviewer: Jeffrey B. Putnam

The study of complex systems has blossomed in the last few years. Many (probably most) interesting non-trivial systems, those that are adaptive or contain a range of complex subsystems, or with nonlinear couplings between components, are well beyond our ability to model using reasonably tractable mathematics. With the rise of computing power, we can now do simulations of complex systems at varying levels and begin to explore their behavior in ways that have never been possible before. Doing such modeling, however, has usually required that one-shot ad hoc programs be built and run to manage the simulation. This book offers an alternative, a way to provide a relatively high-level (and hence abstract) description of interacting complex systems that could be used as a framework for building such models. The framework offers some tools for modeling using traditional mathematics for such things (differential equations and Markov models), but it feels more like a description of a way to build models for simulations. The framework described is not a programming framework (yet), but it is clear that it could be used as the basis for one. This would provide experimenters with the capability to construct a computer-readable description of a system that could be simulated and modified to investigate how changes in the description affect the way the models work. Was it Holland's intent to describe such a framework__?__ Or was the proffered framework intended as a more mathematical tool__?__ Either way, the toolset has a ways to go before we can call it complete. The individual chapters explore different aspects of the framework, such as adaptation, intercommunicating subsystems, isolated subsystems, and languages. These different aspects are usually based on what look like relatively simple (though powerful) computational models, such as classifier systems, grammars, and the like. There is a particular emphasis (as reflected in the title) on systems with boundaries that allow signals to pass through (think of cell membranes or network firewalls) and how these systems interact. The framework can clearly be used to model something, and small examples of such models are provided in a number of places, but can it be used to model (and tractably so, both in terms of description and simulation) larger, more complex systems__?__ It feels that way to me, but lacking a more formal description and an implementation, I'm not sure we can answer that question as yet. The book is an interesting (but often puzzling) read, and requires careful attention, as some of the chapters are concise to the point of being cryptic, and will undoubtedly reveal more on a second or third reading. The text, however, is only rarely complicated or very technical, so the book should be accessible to readers without a technical background or who are not primarily systems researchers. Any real synthesis of meaning for the reader may require more careful thought than is typical. This book is less an answer than a number of interesting questions and potential approaches gathered in one place. Online Computing Reviews Service

Reviewer: Chaim M Scheff

Quite a wondrous epic tale, told in a near-sacred pedagogically pristine "fourth" person, is presented in this book. Thus, it is worthy of achieving a valid academic textbook status, although it is sadly lacking chapter-concluding problem sets. The amazing spectrum-of-knowledge bibliography reads like a list of what I have read and what I still need to read. However, this book claims that the "ultimate goal is to tie ... mechanisms into a single overarching framework that suggests ways to steer complex adaptive systems by modifying signal/boundary hierarchies." In other words, the author contends that there are order-preserving transformations from some endlessly complex systems onto others of similar chaotic order, a metaphor-paradigm that leaves me dumbstruck, instead of inspired. Intellectual ambivalence aside, this is an amazingly well-written book, ripe with sufficient narrative to allow even the most systems-adverse student to appreciate, contemplate, and even enter into the nomenclature and notation of examples from truly diverse disciplines. The book deals with complex systems analysis according to a proper transition of meta-metrics spanning from "adaptive agents (defined by signal-processing rules that provide for parallel processing and adaptation under a genetic algorithm) model[ing] the evolution of hierarchical systems that employ many signals (resources) interacting simultaneously" to "tagged urns (modifications of the urns used in probability theory) [that] use entry and exit conditions to control the flow of balls (signals) between urns, thus providing explicit formal models of semi-permeable boundaries." While this is magnificent enough for the first book of a series, I hope that further volumes will deal with which side of Occam's razor to apply (as noted below). More specifically, the final frontier of the day is systems complexity, which at the moment comes in two types: natural and human-engineered. While the respective metrics of geophysics, meteorology, taxation, and international trade may all be far above any nondeterministic polynomial-time (NP)-complete computational modeling horizon, there is a glaring and significant difference between them. Applying the sharp side of Occam's razor to complex natural systems gives us easier-to-understand models, while applying the dull side to complex human-engineered systems only tells us how unreasonable our human manifestations have become. Natural laws need to be discovered and understood, while human laws need to be re-convolved to allow an ensemble of concurrent self-actualizing progress. Were human systems able to simply evolve, then the survival of the few fittest would be at the amoral expense of the suffering of the many weakest. The author seems to ignore this, as if the laws of human-engineered systems are as immutable as those of natural systems. Even so, he patiently and articulately walks us through a brilliant odyssey of ideas, structures, and formalisms, supported by countless examples that are easy to understand and contemplate. If for no other reason, this pantheon of simply grasped examples makes the book required reading for all of those who may never make the intellectual pilgrimage to all of the exotic worlds cited in the bibliography, even if there may be other, potentially more valid, interpretations for each example. In defensive praise of the author (and in spite of any limitations that I allude to above), the latter half of the text methodically delves into an abstract linguistic meta-dimension of natural systems versus human systems, and specifically the natural and computer language, grammar, ontogeny (really), and mathematical models thereof. Thereafter, the whole kit-and-caboodle is neatly wrapped in adaptive agents and tied with tagged urns, making it a rather suspicious (modeling) object, were it to be left unattended in a public place. While I may never return to the thesis of this work, I do expect to return to the text, to consider the examples, the references, the conceptual orientation, and the intellectual worldview that may ultimately categorize this work as indicative of entangled collections of assumptions that we are still trying to sort out. Thus, I sincerely recommend this book for healthy minds that periodically crave food for thought. My review is honest, if fleeting, praise for the author and for his very in-focus panoramic vision. Online Computing Reviews Service