Expert systems (ES), which are the primary and by far the most widespread application area of artificial intelligence (AI), continue to find new domains to conquer. This book presents six case studies in classical archeology, one in medieval archeology, and one more general study in iconography. These studies are preceded by a 60-page section on general principles and followed by a 28-page epilogue by J. C. Gardin, Director of Studies, Ecole des Hautes Etudes en Sciences Sociales, Paris.
For most computer scientists, the issues discussed in the case studies are largely impenetrable, requiring specialized knowledge of Greek, Roman, Cypriot, Central Asian, and medieval archeology. However, the surrounding chapters, which concern such issues as the motivation underlying the adoption of expert systems in archeology, are very interesting. This discussion is framed within the general context of the application of scientific reasoning to the social sciences. Gardin presents the logicist approach and argues that it is only a short step to ES. He then introduces the SNARK system, an ES shell developed at the University of Paris, which four of the case studies employ.
This book is interesting for its discussion of the logicist program and the deep concern about the application of scientific reasoning to the social sciences. ES as the natural culmination of a reductionist approach to science are carefully analyzed to determine their relevance and potential for making major contributions. Consider the following expression of Professor Gardin’s care in the use of ES:
Our major concern is to give a sounder basis to our constructs in the humanities through methods in which expert systems certainly have their place, but without the computerization of our reasoning ever becoming our main objective (p. 219).
Further, he stresses
Our sole purpose has been to present examples of applications of expert systems in a well-defined domain of the humanities, so as to provide a concrete basis for the discussion of the most interesting questions, raised by this kind of application. . . (p. 224).
Such questions relate to the following epistemological issues: commonsense reasoning, mutual knowledge, the integration of specialized knowledge (field-related logics) into uniform problem solvers, and the relation of such logics to possible-world logics. As a final note of caution, he warns that such concerns do not automatically imply a commitment to logic and reductionism as the only approach to explanation in the social sciences.