According to the preface, this textbook is directed primarily at undergraduate liberal arts colleges and universities, and it should be well-suited for a one-semester or one-quarter introductory AI and LISP course at the second- to fourth-year level. A basic understanding of AI should have a great deal of appeal to liberal arts majors, and since I am all in favor of making science relevant and understandable to everyone who gets a college education, I embarked on the review of this textbook with a great deal of a anticipation. As I went along, though, I found myself reacting quite differently depending on whether I saw the book from the perspective of an instructor or the perspective of a student.

As an instructor, I would like the book quite a bit, except perhaps for the first two chapters; as a student, I would probably find the book challenging and sometimes abstract. I will elaborate later. An issue I soon found myself thinking about, and an issue that affects this as well as most other introductory texts, is the question of whether it is it possible to introduce AI, particularly to liberal arts majors, in a manner that emphasizes the conceptual underpinnings of the science independently of implementation vehicle, or must we require students to grasp the nuances so easily and, for the experienced, so neatly expressed in the code in order to really understand how the ideas hatched in the human brain are really carried out in silicon and electric circuits. This question has no simple answer. Like most things in life, it depends--on the intent and goals of the course; on what we want or expect of the students, both before they enter the first class and after they complete the course; and on the preference of the instructor. My preference would be to emphasize conceptual understanding and to bring in technique to support that understanding. One could argue that LISP (or Prolog, or whatever) is itself part of the AI-related material that needs to be mastered, but I do not agree.

The first two chapters introduce LISP in a bottom-up and prosaic manner; if the students already know some LISP, it would be desirable to go over these chapters quickly or skip them altogether. Even if they do not, it would be a good idea to go through this material as quickly as possible in order to reach the parts both students and instructor are really interested in. Few students will have a desire to learn LISP as it is presented here, in a vacuum with little or no connection to interesting problems. If I were teaching a course from this textbook, I would seriously consider using an auxiliary LISP text; Touretzky [1] would be my favorite.

The rest of the book offers things to like and things to dislike. The things to like include a comprehensive overview of AI topics presented in 16 chapters, three appendices, a glossary, a 15-page bibliography, a name index, and a subject index. In addition, each chapter contains a summary, chapter references, and exercises, most of which are simple. The major benefit here is that students should be encouraged to do some library research and, for a while at least, the textbook could serve as a reference source in the future. The layout is clear and attractive. For the most part, the style is clear although not lively. Not reviewed but also available are an instructor’s guide and an instructor’s diskette that contains all the programs and data given in the textbook, and solutions to the exercises.

The chapter introductory sections are comprehensive and good--for those who already know the topics. In fact, they could serve as exemplary summaries. For the intended audience--students who are encountering the subject for the first time--these summaries are too abstract, packed with too many terms, implications, and nuances that are not always adequately elaborated in the remainder of the chapter.

The book treats all the topics one expects to find in an introductory summary of AI, including searching, knowledge representation, natural language processing, vision, inference, expert systems, robotics, and machine learning. In addition, it contains chapters on neural networks, special and parallel architectures, Prolog and advanced LISP, and social and philosophical issues. One of the appendices gives a simple neural network program implemented in Pascal.

If you are looking for a textbook with wide coverage of the field (which gives you room to customize the course), good references, and LISP, take a look at this one, but be prepared to give your students lots of support and encouragement. If you are not an instructor, you should make sure you have access to a LISP interpreter and enough time to do quite a bit of coding.