The resulting instrument produces interpretations as well as the usual protein electrophoresis curves and component percentages. It is a commercially available productthe first marketed medical device to have used AI techniques in its development.

A major result of this research has been the development of a computer program that is intended to serve as a consultant in a number of medical problem areas. Here the considerations that underlie the program are discussed. The basic functions of the program are outlined in a nontechnical way, and an example of the use of the program is given. Then the results of the use of the program for several different medical problems are reviewed. Finally, an attempt is made to ascertain the potential of programs such as this in the delivery of appropriate medical care. Detailed reports on various aspects of this research are available in the literature (Gorry, 1967; 1968; Gorry and Barnett, 1968a; 1968b), and so the emphasis here will be on providing a general overview of the work and results obtained to date.

Specialization and generalization are accomplished by adding or deleting elements in these lists. The use of symbolic categories of belief (definite, probable, and possible) provides a specifiable means for manipulating the rules. While based on a simple idea, the SEEK program convincingly demonstrates the value of a rich('v structured representation and of reasoning from cases as a way of constructing a model. That is, exjJert knowledge is inseparable from case experience (Schank, 1983), in so far as knov.Jledge explains the cases. The use of a knowledge base to provide an explanatm), model has characterized other recent AIM work as well (cf.

Working in the EMYCIN environment, they were delighted to find that they could more easily distill their criteria for test interpretation by using the production rule formalism.

Every year, as computers become more powerful and less expensive, increasing amounts of health care data are recorded on them. Motivation for collecting data routinely into ambulatory and hospital medical record systems comes from all quarters. Health practitioners require sets of data for clinical management of individual patients. Hospital administrators require them for billing and resource allocation.

Examining the refinement structure created by the automatic programmer makes possible justifications of the code. This chapter describes XP LAIN and outlines additional advantages this approach has for explanation. The significance of Swartout's work is not just its use of a s_vstem design technique that makes explanation possible. His work reveals how principles (here, domain strategies by which specific treatment methods are apphed) are part of explanation. It is useful to supply not just an "audit trail" of what a problem solver did (on perhaps

The work is also of interest because of its relation to psychological studies (Chapter 12) and explanation methodology (Chapter 16).

We have studied difficulties arising in the operations of the "first generation" of AI programs in medicine and have undertaken the development of knowledge representation structures to support needed improvements. The description of a patient in existing programs such as INTERNIST-I (Pople et al., 1975), PIP (see Chapter 6), and MYCIN (Shortliffe, 1976) starts from a single list of findings about the patient. Using a data base of associations between diseases and findings (or rules establishing those connections), these programs form an interpretation of the patient's condition that is essentially a list of possible diseases, ranked by a calculated estimate of likelihood or degree of belief in each. Researchers (Patil, 1979; Pople, 1977; Smith, 1978) have recognized the need to use notions such as causal relationships, temporal patterns, and aggregate disease categories in the description of a program's diagnostic understanding, but the mechanisms provided to do this have been too weak. For example, although causality appears as a term in descriptions in PIP and INTERNIST-I, in both cases its use is limited to guiding the propagation of likelihood measures. These programs fail to capture the human notion that explanation should rest on a chain of cause-effect deduction.

Nor are they claiming that experts do not use general principles for ordering search and selecting alternatives (cf.

Recent research in clinical diagnosis (Barrows et al., 1978; Elstein et al., 1978; McGuire and Bashook, 1978) contributed to a consensus about the general form of the process of clinical diagnostic reasoning. Cues in patient data suggest hypotheses, which are, in turn, tested against subsequent data of the case. The basic hypothetico-deductive process is shared by experienced and inexperienced diagnosticians alike, as are numerous parametric characteristics of the process, such as the percentage of data items to first hypotheses, the average number of hypotheses maintained in active consideration, etc. These studies, however, have generally neglected the content of diagnostic reasoning, that is, the knowledge base of medical subject matter involved in the diagnostic process. Yet, despite prevalent findings of lack of differences in the form of diagnostic reasoning as a function of experience, the few differential findings from these research efforts implicate the importance of the knowledge base.