Introduction by the proud father, Jean-Luc Marcé:
Expert systems (or rule-based decision management systems, as they are often known) have been a mainstay of computer science for decades, and will continue to dominate the field of knowledge automation. They are actively being taught in schools, as illustrated by this paper, first of two written by my daughter Sanjana, an 11th grade student at the Harker School. This paper covers the history of expert systems from the 1950’s to present time.
What is an Expert System? by Sanjana Marcé
Since its advent, technology serves the purpose of facilitating, accelerating, and ultimately replacing the work of humans and is designed to complete the same tasks with less man labor and in less time. The field of artificial intelligence, a relatively new take on the goal of technology, proposes the development of software with the capabilities of human sensibility and reasoning. The conce
pt originated as early as 1950 with the invention of the Turing Test, mathematician Alan Turing’s method of quantifying the intelligence of a computer system by an individual’s ability to distinguish between the machine and a real human being_ A subset of AI, expert systems attempt to recreate the expertise of live human professionals within the framework of computer software, designed to help the user solve a certain, defined problem, referred to as the task domain of the expert system, efficiently and effectively._ Although not as popular as during the 1970s and 80s, expert systems are still widely implemented in industry, government, commerce, and personal use.
Expert systems imitate the decision making process and reasoning functionalities of professionals by combining two essential components: the knowledge-base and the inference engine. The knowledge-base portion of the expert system is the translation of the human expert’s information reserves — both factual, accepted knowledge and heuristic, experiential proficiency developed through practice._ Knowledge engineers act as the link between the human expert and the electronic expert system, transferring the professional’s information base to the machine._ Meanwhile, the inference engine is the algorithm of the program that decides what conclusions or decisions to make based on the asserted facts and the sets of assigned rules._ The ability of the problem-solving paradigm to chain rules together transforms the program from a set of conditions and actions to a pathway of reasoning eventually leading to a conclusion. Often, the task domain at hand presents itself to the user in two ways: either a set of conditions are known and a decision must be made from the given scenario, or a conclusion or goal has been outlined and a hypothesis is to be generated. From this dilemma arose the concepts of forward and backward chaining._ Whereas in forward chaining an initial state presented to the expert system causes a series of rules to be evaluated until the program arrives at a decision, in backward chaining the end condition is known and the program is used to form a plausible hypothesis._ As can be imagined to avoid conflicting reasoning paths, complex sets of rules require precedence or weight to be assigned. This method, which helps establish the level of certainty of the expert system’s predictions and recommendations, is referred to as the confidence factor. A program line of reasoning based on uncertain information is sometimes noted as fuzzy logic._ The capability of fuzzy logic and confidence factors expands the breadth of expert systems given that hypotheses are also valid conclusions of the software. Additionally, some expert systems provide the user with an explanation for the arrived conclusion, often by tracing the inference engine’s rule evaluations._ The ability for the expert system to formulate a best guess based on the available information further likens it to actual human experts who also must make calls of judgement on occasion.
Yet the process and implementation of expert systems were not always so refined. Rather, only until the origin of the LISt Processing language, or LISP, in 1958 out of MIT was a new platform created for list-based programming solutions. Fifteen years later, and across the Atlantic Ocean, French researchers launched Prolog, or PROgramming in LOGic, which first used facts or assertions to represent information and used rules to operate on this information. Yet the success of Prolog would not have been possible without the Japanese pioneering the Fifth Generation Computing Systems Project during the 1980s, which made aware the necessity of Prolog as a logic-based programming language._ In this way, although most of the original research leading to the development of expert systems started within major universities of the United States, the process would not have been possible without the global effort. Additionally, the application of expert systems to the commercial world as soon as the idea arose further convinced commercial and private users of the power of expert systems. The first example of this was Dendral, an expert system developed in the 1960s and used in conjunction with the work of chemists identifying compounds in samples through mass spectroscopy. Created at Stanford under the guidance of Edward Feigenbaum, popularly known as “the father of expert systems”, Dendral automated the decision making process of mass spectroscopy and improved the accuracy of chemists’ research._ Other institutions, upon witnessing the success of Dendral in the chemical field, rushed to participate in the budding new concept of expert systems. MYCIN, headed by Edward Shortliffe at Stanford during the 1970s, tackled the diagnosis of bacterial infections and appropriate treatment for patients._ Perhaps one of MYCIN’s most important contributions to the development of AI and expert systems was the isolation of the knowledge-base from the inference engine with the creation of EMYCIN. EMYCIN was the first expert-system shell, acting as a reusable model for logic-based programming with only the raw data and rule set to be inputted._ Over the next decade, expert systems expanded to the environmental industry with PROSPECTOR, a program designed for mineral analysis; the oil industry with DIPMETER, a software for oil data analysis; the technical world with XCON, developed 1978 by John McDermott at CMU which verified that consumers received all necessary parts for their desired task; and the space exploration industry with CLIPS, NASA’s own expert system adapted to organizing space shuttle launch schedules._ During the same time period, Dr Charles L. Forgy of CMU pioneered the Rete, or “net”, algorithm, providing expert system developers with a more efficient means of running the inference engine. With the new technique, all asserted facts would first be compared to a large net-like structure of conditions checked by the system, matched functions would be carried out, and only those facts changed by the previous operations would be reprocessed through the condition net._ The new method, still used today, provided a much faster means for the expert system to arrive at a decision from the asserted facts.
Since expert systems represent a method of connecting vast sources of data to logical conclusions and decisions, they are often used in modern corporations and organizations for medical and electronic diagnoses, planning and scheduling, configuration verifications, and finance. With time, expert systems have also shifted to be integrable with modern programming languages as opposed to expert system-specific ones like LISP and Prolog. However, expert systems remain exposed to the difficulty of managing large database sizes, the struggle of handling conflicting rules, and the question of liability in the event of incorrect or harmful decision-making. While not as widely studied as in the 1980s, expert systems can still be found in everything from agricultural production, to industrial factory inspections, to educational tools like Mathematica, to personal use tools like grammar checkers and electronic tax-help software.
Belleguelle, Steve. "Expert Systems." AIM Expert Systems. AIM, n.d. Web. 4 Feb. 2016. Link
Bullinaria, John A. "What Is an Expert System?" University of Birmingham School of Computer Science. U of Birmingham, 2005. Web. 4 Feb. 2016. Link
"Expert Systems and Applied Artificial Intelligence." University of Missouri–St. Louis. U of Missouri–St. Louis, n.d. Web. 4 Feb. 2016. Link
Feigenbaum, Edward. "Knowledge-Based Systems in Japan." Japanese Technology Evaluation Center. Ed. Robert S. Engelmore. WTEC Hyper-Librarian, May 1993. Web. 4 Feb. 2016. Link