The Intent of the protocol is to show AGE from the user's point of view. The main body describes the motivations for and the description of the facilities in AGE.

AGE is a system designed to aid knowledge engineers build knowledge-based programs. We have built a laboratory of building blocks which the user can assemble in various ways to fit a particular problem. This paper describes the facilities available to the user and descibes an example program built with AGE.

A clinical investigator graphical capabilities which can plot specific parameters for a keeping the records of his study patients on such a system can patient over time 1126]. However, it is in the analysis of stored use the program's statistical capabilities for data analysis.

A program for selecting antibiotic therapy is described that makes sophisticated recommendations and provides simple, useful explanations of its reasoning for a user. Our method is to structure the therapy optimization problem in terms of local and global solution criteria that are applied in a generate-and-test algorithm. Generation of therapy recommendations is directed by a fixed, ordered set of canonical instructions that describe the global characteristics of a recommendation. Other factors are dealt with in a "planning" phase particular to each organism for which therapy is to be prescribed, and the test phase that incorporates patient considerations such as allergies and age. We demonstrate the advantages of this canonical form for explanation, comparison of alternative recommendations, and a simple instructive capability.

These knowledge bases are generally built by interviewing human experts to extract the knowledge they use to solve problems in their area of expertise. However, it is not clear that the organization and level of abstraction of this performance knowledge is suitable for use in a tutorial program. We are exploring this problem in the GUIDON tutorial program, using the knowledge bases of MYCIN-like expert systems. MYCIN is a knowledge-based program that provides consultations about infectious disease diagnosis and therapy (Shortliffe, 1974). In MN CIN, domain relations and facts take the form of rules about what to do in a given circumstance. A principle feature of this formalism is the separation of the knowledge base from the interpreter for applying it. This makes the knowledge accessible for multiple uses, including application to particular problems (i.e. for "performance") and explanation of reasoning (Davis, 1976). We have most recently used the MYC1N knowledge base as the foundation of a tutorial system, called GUIDON.

We describe the development and (partial) Implementation of an "automated consultant" to advise non-expert engineers In the use of a general-purpose structural analysis program. The analysis program numerically simulates the behavior of a physical structure subjected to various mechanical loading conditions. The automated consultant, called SACON (Structural Analysis CONsultant), Is based on a version of the MYCIN program [Shortliffe74], originally developed to advise physicians In the diagnosis and treatment of infectious diseases. The domain-specific knowledge in MYCIN Is represented as situation-action rules, and is kept independent of the "inference engine" that uses the rules. By substituting structural engineering knowledge for the medical knowledge, the program was converted easily from the domain of Infectious diseases to the domain of structural analysis.

The prescriptions were placed in rat dom Materials and Methods CASE 6.--A 42-year-old woman had a order and in a standardized format to Ten patients with infectious meningitis three-day history of headache, nausea, and disguise the identities of the individual were selected by a physician who was not fever.

PUFF is now in routine use in Presbyterian Hospital, Pacific Medical Center (PMC), in San Francisco. The program produces a report, intended for patient records, that explains the clinical significance of measured quantitative test results and gives a diagnosis of the presence and severity of pulmonary disease in terms of the measured data, referral diagnosis, and patient history. "Rules", or statements of the form "IF condition THEN conclusion ", are used by the physiologist and the computer system to specify the system operation. The sequence of rules used to interpret the case also specifies a line of reasoning about the case, or the detailed explanation of the interpretation of the case. The use of rules for this type of knowledge based system is taken from the results of applied Artificial Intelligence research. In a 144 case prospective evaluation, there was a 91% overall rate of agreement between the rule based system diagnoses and the diagnoses of the designing physiologist; there was a 89% rate of agreement between the system diagnoses and diagnoses of a second independent physiologist.

References 67 October 27, 1977 1 Introduction This application addresses the continuation of research on the applications of artificial intelligence (Al) (1) to experimental molecular genetics. It is an extension of a longstanding effort to cultivate attention to ongoing laboratory research as a domain of explorations in artificial intelligence. Our major effort in this field had been in the DENDRNL project, with analytical organic chemistry as the object discipline.

The performance of a computer-based clinical consultation system is evaluated. The program, called MYCIN, is designed to function as an aid for infectious disease diagnosis are therapy selection, with an initial emphasis on bacteremias. The evaluation methodology is discussed, as well as the difficulties encountered in attempting to evaluate clinical judgments. Specialists in infectious diseases judged MYCIN's final therapy recommendation, and intermediate conclusions about the significance of the infection and identity of infecting organisms. The evaluation techniques described may be useful in assessing the performance of other clinical decision aids. Results of the evaluation show that the program's therapy recommendations meet Stanford experts' standards of acceptable practice 90.9% of the time (Table II), with some variation noted both among individual experts and between Stanford experts and others (Tables I and II).