Browse Books

Reviewer: H. Van Dyke Parunak

The road from the research laboratory to the daily life of the common citizen is long and treacherous. Many brilliant technical ideas never affect anyone beyond the research community; many more are delayed for years before they become commonly known and used. Resnick's technical domain, artificial life (A-life), is a new traveler on this road, and his work focuses not on discovering new ideas in this domain, but on how concepts of decentralization and emergent behavior can be made understandable to the public at large. Resnick's book grew out of his work at MIT with Hal Abelson and Seymour Papert. In a previous technical generation, Abelson and Papert developed and used the Logo language as a sandbox to introduce students of all ages to computers, von Neumann computation, and differential geometry. Resnick follows in this fruitful tradition by devising a massively parallel dialect of Logo to serve as a sandbox for decentralized systems with emergent properties. His dialect, StarLogo, was originally developed on the parallel Connection Machine, and while versions for other computational platforms are under development, at present people outside MIT can use the language only on a Connection Machine. The book reports not only the general nature of the language and some particular simulations that it supports, but also the experience of high school students in working with it and in coming to grips with the new ideas it embodies. Chapter 1, “Foundations,” discusses the wide range of domains in which decentralized ideas are becoming prominent. Centralized control is increasingly out of favor in organizations. The emphasis in computer technology has shifted from the centralized mainframe to decentralized networks of personal processors. Scientific models in fields as diverse as materials science and ecology are increasingly decentralized, and students of the human mind and of the nature of knowledge also draw on these concepts. Chapter 2, “Constructions,” confronts the difficulty people have in understanding decentralized systems, and expounds Papert's constructionist approach to education. In this approach, learners construct knowledge from their experiences in the world, by actively creating something that is personally meaningful to them. MIT's LEGO/Logo project exemplifies this approach, but does not easily support the large number of entities required for A-life studies. StarLogo was devised to meet this need. StarLogo reifies not only large populations of turtles, but also (through “patches”) the environment in which they move. A central A-life tenet is that the environment is active, not passive, and interacts with its inhabitants in complex ways. Patches can do almost anything a turtle can do, except move. In the examples discussed in the book, their most common use is to model the presence and diffusion of pheromones in insect societies. Chapter 3, “Explorations” (the longest chapter) describes nine experiments in the StarLogo world, with complete StarLogo listings and a detailed discussion not only of the programs but also of the experience of students in thinking their way through the problems. Six of the simulations are of naturally observed situations: how individual slime mold cells aggregate into the mold's aggregate phase; how ants form paths to food sources; how traffic jams form; how termites coordinate their deposition of waste to build termite hills; how the populations of predators and prey depend on one another; and how the spread of a forest fire depends on the density of trees . The other three experiments emphasize that StarLogo's main purpose is education in decentralized mechanisms, not research simulation. “Turtles and Frogs” models the tendency of mixed populations to segregate to a far greater degree than individual members desire. Conventional Logo has long been used to teach geometry, and “New Turtle Geometry” illustrates how massive parallelism gives the subject a new twist (for example, drawing a circle by turning loose a large population of turtles with random headings from a single point, or graphing a function with the aid of smart patches, each of which changes color to indicate whether it is on the function). “Recursive Trees” shows how another classic application of Logo becomes simpler in the highly parallel world. Chapter 4, “Reflections,” generalizes from the experience of Resnick's students to reflect on the natural persistence of the centralized mindset and suggest several heuristics for achieving a more decentralized view of the world. Chapter 5, “Projections,” briefly anticipates how tools like StarLogo can help students toward this end. Appendix A, “Student Participants,” describes the student population and experimental setting of the exercises. Appendix B, “StarLogo Overview,” summarizes the major features of StarLogo, but is not a reference manual for the language. Resnick writes clearly and persuasively. Even without access to a StarLogo implementation, the book will help readers understand decentralized emergent systems. Unfortunately, at present this vicarious experience is the only benefit most readers will enjoy, since implementations of StarLogo for common platforms are not readily available. This volume may be just the incentive some creative hacker needs to fill this gap, so that the rest of us can enjoy first-hand the discoveries Resnick's high school students experienced.