Expert systems are a particular type of Knowledge Based Systems. Their name comes from two sources: one, that “they are able to learn by themselves” and the other, that the knowledge they manage comes from expert’s consultation. (Tyler) Typical applications of expert systems include tasks such as medical diagnosis, fault localization equipment and interpretation of measurement results. The expert system needs to solve problems that require for their solution expertise in some specific area. Thus, the expert system should have this knowledge in one form or another. Therefore, they are also called systems based on knowledge. However, not every system based on knowledge can be seen as the expert.
The expert systems should also be able to explain their behaviour and their decisions in some way to the user, as well as does the expert. This is necessary in areas which are characterized by uncertainty or information asymmetry (e.g. medical diagnosis). In these cases, the ability to explanation is needed in order to increase the user's confidence to the advice system, and in order to allow the user to detect a possible defect in the system. In this regard, the expert system should provide user-friendly interaction with the user, which makes the reasoning process of the system "transparent" for him.
Expert systems have certain advantages. In particular, expert system:
exceeds the capabilities of humans in solving extremely cumbersome problems;
has no preconceived ideas, while the expert uses side knowledge and is easily influenced by external factors;
does not jump to conclusions, neglecting certain phases of the withdrawal;
provides a dialog mode;
allows working with information that contains the symbolic variables;
provides the correct work with information that contains errors due to the use of probabilistic methods of research;
allows simultaneous processing of alternative versions;
explains the implementation steps of the program on demand;
provides the ability to justify decisions and play the way of its adoption.
However, even the best of the existing expert systems have certain limitations compared to a human expert, and they are as follows:
Most expert systems are not quite suitable for wide use. If the user does not have some experience with these systems, it can cause serious difficulties for him. Many expert systems are available only to those experts who have established their knowledge base. Therefore, it is necessary to develop the appropriate user interface that would provide the end user inherent operation.
The skills of expert systems do not always increase after a session of the examination, even when they manifest new knowledge.
Bringing knowledge from the expert into a form that would ensure their effective use still remains a challenge.
Expert systems typically cannot acquire qualitatively new knowledge, not provided for in the design, and especially do not possess common sense. Man is an expert at solving problems, who usually appeals to intuition or common sense if there are no formal methods of solution or equivalent solution to this problem.
The main difference between expert systems and other software tools is a knowledge base in which knowledge is stored as entries in a knowledge representation language that allows you to easily change and add to the knowledge base in a form that is understandable for experts — developers of expert systems. In usual programs, the knowledge is hardwired into the algorithm and only the author can edit them (if he remembers the program). Expert systems are important in areas where empirical knowledge dominates and the accumulation of facts has outstripped the development of theory (Medicine, Geology, Finance, etc.). Such well-structured areas as mathematics, physics, theoretical mechanics, are based on the developed mathematical apparatus for the description of its regularities, which allows conducting computer simulations using traditional algorithmic programming (without selecting level of knowledge). (García Bringas, Hameurlain and Quirchmayr) Expert systems are important where the definitions are blurred, concepts are changing, and the situation depends on many contexts, where there is great uncertainty, the vagueness of the information.
In my humble opinion, the road to successful and applicable expert systems is long and difficult. For example, in medicine very few successes have been reported. Even for successful systems such as MYCIN, ONCOCIN and AI-RHEUM there is as yet no widespread market. This is also true for INTERNIST, developed by Myers and Pople, and its successor program CADUCEUS. Some systems, however, have direct spin-offs, such as the QMR program (Quick Medical reference), based on INTERNIST. (Rienhoff, Piccolo and Schneider)
We may wonder, now that the first period of enthusiasm is over, in what ways are expert systems different from our earlier decision models. In a very pessimistic mood, we might conclude that expert systems are not useful for experts, and that they will not be used by, and usable for, non-experts. Such discouragement, however, is unwarranted. On the contrary, we should, based on the lessons learned from the past, look for new directions for our research and strive for practical projects.
All in all, I believe that expert systems will make a huge contribution in our future. For now there are thousands developed highly specialized expert systems. This suggests that the expert systems constitute a very significant part of the software.
Works Cited
García Bringas, Pablo, Abdelkader Hameurlain, and Gerald Quirchmayr. Database And Expert Systems Applications. Berlin: Springer, 2010. Print.
Rienhoff, O, U Piccolo, and B Schneider. Expert Systems And Decision Support In Medicine. Spinger, 1988. Print.
Tyler, A. R. Expert Systems Research Trends. New York: Nova Science Publishers, 2007. Print.