In the early days of computing, these goals were central to the new discipline called cybernetics [126], [2]. Over the past two decades, progress toward these goals has come from a variety of fields - notably computer science, psychology, adaptive control theory, pattern recognition, and philosophy. Substantial progress has been made in developing techniques for machine learning in highly restricted environments.

Accepted February The author is grateful to C. Djerassi for encouragement to 13, 1978. Work supported by the National Institutes of Health publish these findings and for reading the manuscript.

Success of a knowledge-based program depends on both competence and acceptability. It must perform well for it to be worth using, but is must be acceptable to users for it to be used. There are many dimensions to developing competent and acceptable knowledge based systems which can serve as "intelligent af.sistants-for problem solvers in science (see Shortliffe and Davis, 1975). One of these is the old Al problem of representation of knowledge. Since most previous work on representation has stressed its importance for problem-solving (e.g.

Reprinted with permission from Analytical Chemistry, August 1976, Vol.48, No.9, pp.1368-1375.

Over tt7e years a range of different mechanisms have been proposed and used (e.g., standard procedure invocation, goal-directed invocation, etc.), each typically motivated by the attempt to develop new forms of knowledge encoding (e.g., procedures, PLANNER theorems, etc.). We consider in this paper tne strengths and weaknesses of a range of these mechanisms, paying particular attention to their expressiveness and validity. This analysis brings to light certain shortcomings shared to some degree by all current mechanisms. A number of ideas are presented as the basis for a mechanism which appears to offer a way of overcoming the problems discovered. We describe how those ideas have been implemented and tested in a rule-based system, and explore their impact on system performance, ease of construction, and flexibility. We consider also their value as a generalization of the existing notions of procedure calling. Though the terminology may differ, some of the shortcomings we point out and some of the ideas proposed may be recognized by others who have built similar systems, where some of these ideas have been Implemented in various Informal ways. The purpose of this paper is not, therefore, to advocate a particular solution, but instez.d

COMPUTER PROGRAM APPLIED TO INFECTIOUS DISEASES* Edward H. Shortliffe Cepartmenc of!'edicine Stanford University School of Medicine Stanford, California 94305 A rule-based expert system is described which uees artificial intellieence techniques, and a model of:he iateractica between phesiciane and human consul-:ants, to attempt to satisfy the demands of a user:o unIr7 that is often reluctant to experiment with touter zecnnology. Experteace to date has demonstrated that the program is efficient, relacively easy to use, and reliable in the domain ofbacearemei therapy selection. Future work will involve broadening dad evaluatim; tne program's expertise in other areas of infectaoue disease therapy. Ihtreductioa Few eotentialusereopulations are as demanding of tomeuter tecenology as are practicing physicians. This our to a variety of factors which include the,eysecian's independeace as a lone decision maker, the seriousaess wieh which he views actions that may often have life-and-Ceach sigaifizance, and the overwhelming:t.me

The low cost of microprocessors today, and the future trend in both cost and performence, makes large microprocessor-networks very interesting. A net of a thousend processors or more can be built using present techniques. However, there are several problems in utilizing such a horee of processors in a rcesonebly efficient way. There also seem to be restrictions to the kinds of opplications that can be mapped on to such a system. One control mechanism that seems to be very useful, at least for some Artificial Intelligence type problems, is the CONTRACT NET, ISmith77]. In this paper we will first look at some of the desirable characteristics of a lerge multi-microprocessor net. Than we will describe several different organizations together with their advantages and problems. We are also discussing broadcasting in lattices using circuit switched minimum spanning trees.

A rule-based expert system is described which uses artificial intelligence techniques, and a model of the Interaction between physicians and human consultants, to attempt to satisfy the demands of a user community that Is often reluctant to experiment with computer technology. Experience to date has demonstrated that the program Is efficient, relatively easy to use, and reliable in the domain of bacteremia therapy selection. Future work will involve broadening and evaluating the program's expertise In other areas of infectious disease thPrapy. To that end rules regarding diagnosis and treatment of meningitis have been written and are currently under evaluation.

The general practitioner has accordingly become rare, and today's primary care physicians are beginning to graduate from family practice residencies which recognize that "family doctoring" is a subspecialty in itself. Thus when a patient's problem clearly falls outside the area of the attending physician's expertise, consultations from experts in other subspecialties have become a wellaccepted part of medical practice. Such consultations are acceptable to doctors in pert because they maintain the primary physician's role as ultimate decision maker. The consultation generally involves a dialog between the two physicians, with the expert explaining the basis for his advice and the nonexpert seeking jus-MYCIN Project is located at Stanfcrd University School of Medicine and is zepported by BHSRE Grant No. HS01544. Much of the work described in this report yes undertaken by other project members, notably A.C. Scott and W.J. Clancey, who have devoted much of their time to improvements in the general question-answerer, and R. Davis, who did most of the work on the reasoning status checker and on knowledge acquisition capabilities.

A new predicate calculus deduction system based on production rules is proposed. The system combines several developments in Artificial Intelligence and Automatic Theorem Proving research including the use of domain-specific inference rules and separate mechanisms for forward and backward reasoning. It has a clean separation between the data base, the production rules, and the control system. Goals and subgoals are maintained in an AND/OR tree to represent assertions. The production rules modify these structures until they "connect" in a fashion that proves the goal theorem. Unlike some previous systems that used production rules, ours is not limited to rules In Horn Clause form. Unlike previous PLANNER-like systems, ours can handle the full range of predicate calculus expressions including those with quantified variables, disjunctions and negations.