In his new book Alternate Realities: Mathematical Models of Nature and Man (New York: John Wiley and Sons, 1989, 493 pages, $34.95), John L. Casti gives us an impressive, up-todate look at several areas of mathematics that are being applied to the study of biological and sociological systems. These areas, including cellular automata theory, catastrophe theory, nonlinear dynamics and chaos, game theory, and control theory, are finding use on the frontiers of scientific research. Although these areas and their applications are described in various other sources, both on the level of a scientist and a layperson, I know of no other book that brings them all together to show how they can be used in scientific research. However, this book suffers from being written for mathematical specialists and, therefore, limits the potential readership. An opportunity to educate more scientists in the use of mathematical models is regrettably missed.

This issue of the AI Magazine initiates a new and (we hope) regular feature, Reviews of Books. Before presenting our first book review, a few comments about the aims of this feature are in order. Visions of applications of computer technology can serve as a stimulus for discussions of new ideas and directions in this field. Consequently, in addition to reviewing the standard literature produced by those in the AI community, we also seek to encourage reviews and discussions by people outside the field of AI to characterize where we seem to be going and what we are likely to achieve in the future. Reviews of this type can focus on such things as the technical accuracy of these popular treatments, and/or moral conclusions, either explicit or implied.

This article is a review of Thinking about Android Epistemology by Kenneth Ford, Patrick Hayes, and Clark Glymour. The result of colliding beams of gold traveling near the speed of light ("minibangs") allows physicists to observe the liberation of quarks and gluons from protons and neutrons, revealing conditions that existed at the earliest moments of creation of the universe, thus validating current theories of how the original mix of quarks and gluons phase-transitioned into the mundane soup of protons and neutrons that forms the building blocks of everything. Theoretical and experimental breakthroughs since the 1970s, as well as technological advances in the art of colliding and detecting particles, have made it possible to observe a new "energy frontier," with a wealth of results that will allow a refinement of our theories. The question of the validity of the results obtained is a completely empirical matter. No one would seriously entertain the claim that the results obtained from RHIC are invalid because the results were obtained in an artificially induced laboratory setting rather than as the result of direct observation of nature.

Intentions in Communication is the outgrowth of an interdisciplinary workshop on the role of intention in theories of communication. Attending the workshop were researchers in computer science, linguistics, philosophy, and psychology. The resulting book contains edited versions of 14 papers (13 of which were presented at the workshop), commentaries, and an introduction. The topics of these papers range from philosophical analyses of the concept of intention to algorithms for recognizing plans, from logical formalizations of speech acts to analyses of intonational contours in discourse. The idea of relating intentions to the use of language is an outgrowth of speech act theory.

As Hofstadter points out, analogies are fluid, meaning that the analogy between two entities can be drawn differently depending on how these entities are represented. The analogy that is drawn, in turn, can change the representation of the entities being compared. Thus, the analogy between Hofstadter and Sagan can be seen as positive: Both have explained important concepts in their fields to a wide audience and transmitted the excitement of these ideas. Both have inspired a number of people within their fields. Unfortunately, a more negative analogy between Sagan and Hofstadter is possible.

Early research on AI typically involved qualitative demonstrations of intelligent behavior, with novelty being the primary focus. However, as the field has matured, there have been increasing demands for more careful evaluation using quantitative measures of behavior. In some cases, the response has taken the guise of formal analyses, and in others, it has emphasized comparisons between system and human behavior, but the predominant movement has been toward empirical studies of AI methods. As a result, techniques for experimental design, exploratory data analysis, and statistical testing, originally developed in other fields, have become increasingly relevant for AI researchers. Paul Cohen's book Empirical Methods for Artificial Intelligence aims to encourage this trend by providing AI practitioners with the knowledge and tools needed for careful empirical evaluation.

After outstanding service as Book Review Editor, B. Chandrasekaran has completed his term. His energetic guidance of the book reviews section brought the magazine a rich set of reviews that were always eagerly anticipated, and I would like to thank him for his important contributions. I am delighted to announce that Michael Wellman, of the University of Michigan, has agreed to join the magazine as the new Book Review Editor. I know that the book reviews feature will flourish under his stewardship, and I enthusiastically welcome him to the magazine. I would also like to draw readers' attention to a related addition to the AI Magazine web site (www.aimagazine.org).

Remember that time and space are a priori conditions of human perception in Kant's philosophy. On the one hand, time is inherent to action and change; on the other, action and change are possible because of the passage of time. According to McDermott, "Dealing with time correctly would change everything in an AI program" (McDermott 1982, p. 101). It should not be surprising then that temporal reasoning has always been a very important topic in many fields of AI, particularly areas dealing with change, causality, and action (planning, diagnosis, natural language understanding, and so on). AI developments based on temporal reasoning lead to general theories about time and action, such as McDermott's (1982) temporal logic, Vilain's (1982) theory of time, and Allen's (1984) theory of action and time. Work on the application of these results has taken place in fields such as planning and medical knowledge-based systems. However, action and change are not an exclusive interest of AI.

ATP systems are used in a wide variety of domains: A mathematician might use the axioms of group theory to prove the conjecture that groups of order two are commutative; a management consultant might formulate axioms that describe how organizations grow and interact and, from these axioms, prove that organizational death rates decrease with age; or a frustrated teenager might formulate the jumbled faces of a Rubik's cube as a conjecture and prove, from axioms that describe legal changes to the cube's configuration, that the cube can be rearranged to the solution state. All these tasks can be performed by an ATP system, given an appropriate formulation of the problem as axioms, hypotheses, and a conjecture. Most commonly, ATP systems are embedded as components of larger, more complex software systems, and in this context, the ATP systems are required to autonomously solve subproblems that are generated by the overall system. To build a useful ATP system, several issues have to ...

The eight sections are (1) Artificial Intelligence (introductory material); (2) Problem-Solving (search and game playing); (3) Knowledge and Reasoning (propositional and predicate logic, inference techniques, knowledge representation); (4) Acting Logically (planning); (5) Uncertain Knowledge and Reasoning (probabilistic reasoning, Bayesian nets, decision-theoretic techniques); (6) Learning (inductive learning, neural nets, reinforcement learning); (7) Communicating, Perceiving, and Acting (natural language processing, computer vision, robotics); and (8) Conclusions (philosophical foundations and summary). What makes this textbook so good? First, it is remarkably comprehensive. In the preface, the authors suggest several alternative paths through the book that could serve as the basis of a one-semester course. At the University of Pittsburgh, my colleagues and I cover roughly the first half of the book (Sections 1-4) in the firstsemester introductory graduate AI course, covering most of Sections 5 through 8 in a second-semester course.