Project
Paper
Melinda Smith -
c203111
When looking for a solution to any problem, an expert is generally preferred over a novice. Imagine being able to consult with an expert in a particular field whenever necessary. Think of the time and resources that could be saved. This entity clearly could not be a human, based on it’s availability alone. This entity would be an expert system. What, then, is an expert system?
An expert systems (ES) can be defined in various ways. If it is defined based on what they do, one could say an Expert System solves problems efficiently and effectively in a narrow subject field. If it is defined based on how they work, they would be described as an intelligent computer program that uses knowledge and inference procedures. A more thorough definition would incorporate both the functional and structural aspects of these definitions. Thus an expert system is a computer application program that takes the knowledge of one or more human experts in a field and computerizes it so that it is readily available for use. It contains acquired expert knowledge or knowledge base and uses inference rules to imitate the expert’s evaluation process in order to offer a conclusion.
Expert systems usually contain two components; the knowledge base and the inference engine.
The Knowledge Base:
The knowledge base houses the knowledge of one or more human experts in a specific field or task and is usually made up of factual knowledge, information that is commonly shared, usually found in textbooks or journals, and typically agreed upon by humans knowledgeable in a specific field or task. Sometimes the knowledge base may include heuristic knowledge, which is experiential knowledge of performance; the knowledge behind “an educated guess.” The knowledge base is set up as an “intelligent” database in that it can usually manipulate the stored information in a logical, natural, or easy-to-find way. Because of the way it is set up, it can conduct searches based on traditional data search techniques , as well as predetermined rules of defined associations and relationships. Programming in the knowledge base is formatted as an IF ... THEN logical rules structure. This structure is a series of IF conditions that, if met, THEN a specific result may be concluded.
The Inference
Engine:
The inference engine usually is setup up to mimic the reasoning, or problem solving abilities that a human expert would use to arrive at a conclusion. This component analyzes and processes the rules in the knowledge base. The inference engine determines which rule to look at next and conducts searches based on heuristic information in order to minimize problem solving time. It simulates the evaluation process of relating the information and rules in the knowledge base to the answers to a series of questions given by the operator or user.
Expert Systems are created by Knowledge Engineers, who interview experts in whatever field is being considered. The Knowledge Engineers extract the essential skills from the experts and codes them as rules. The inference engine software, which can be purchased or downloaded from the internet, is interfaced with the knowledge base. When the Expert System is executed, it usually asks the user a series of questions to try to steer the user to find patterns that the expert system can apply to its database of knowledge, attempting to narrow down the potential answers until the system can make an educated guess at a final answer. Presently, expert systems can be bought off-the-shelf for common applications or purchased customized by any number of software houses.
The earliest expert systems can be traced back as far as 1972. DENDRAL was the first expert system and thus created the field of knowledge engineering. It was set up to interpret mass spectra on organic chemical compounds.
MYCIN was designed by Stanford
University scientists to provide
consultative advice on diagnosis and therapy for infectious diseases. Later a natural language interface, called
TEIRESIAS, was added so that the program could be questioned about it’s
conclusions. In fact many of the early
expert systems were concerned with the field of medicine because medicine is ideal for demonstrating
the contributions of this AI sub-field.
MACSYMA—is a symbolic mathematics
expert system that does math problems involving algebra, calculus, or
differential equations. It represents the first success of an old style of
knowledge-based systems. It uses its processing speed and exhaustive searching
techniques to arrive at solutions.
MOLGEN—was set
up to demonstrate the object-oriented representation of knowledge- based expert
systems for the planning of gene-cloning experiments in the laboratory.
XCON—is an expert system used by DEC to configure, or set up, VAX computers.
These are large computer systems bought in modules (processor, memory, disk
drives, terminals, etc.) that are configured for specific tasks based upon the
assembled modules. When it first was introduced, XCON, contained 2500 rules and
could handle computer system setups involving 100-200 modules.
INTERNIST—is the largest expert system used in the medical profession and is
growing every day. It will ultimately be a diagnostic tool for the entire field
of internal medicine. Only 25% of the field has been developed in the knowledge
base in the past 25 years. Diseases, symptoms, etc. are coded as objects and
are allowed to interact according to a series of rules.
PROSPECTOR—is an ore/mineral deposit locator and was the first to use semantic
networks in its knowledge base.
PUFF—is a medical diagnostic tool used to interpret the results of a
respiratory test. It includes an expert system shell based on the MYCIN model
called EMYCIN. While MYCIN has taken 50 person-years to complete, PUFF has
taken only 5 person-years.
Presently, expert systems are being used to aid human decision making in many areas beside medicine and science. Accounting has numerous expert systems which it uses for assessment of various problems. For instance, Planet, is a rule based expert system, that performs strategic and detailed audit planning. Employment search is another arena in which expert systems are being used to aid in human decision making. The Employment Coach is a knowledge based decision support system to help people find and keep work. Even the field of agriculture has adopted the use of expert systems. The Soy Trader is a knowledge based decision support system for participants in Soybean Markets.
Expert systems offer many advantages to it’s users. Expert systems do not forget. They can be reproduced thus making their expertise available to a greater number of people and thereby saving time and money. Expert Systems are consistent in their advice giving, since they are un-influenced by most recent events or early information. Expert systems can provide documentation of the decision process, thus giving the user tangible proof of how a decision was reached. They save time since, information is available sooner for decision making. And the greatest advantage, in my opinion, is that a single expert system can combine the knowledge of multiple experts thus giving it more “breadth “, that is a greater scope of knowledge, than a single person is likely to achieve.
While expert systems boast of some impressive advantages, it’s disadvantages should not be overlooked. For example, expert systems lack common sense and it is not yet known how to give them common sense. They also lack creativity, i.e. they cannot respond creatively to unusual situations. Expert systems that have only a knowledge base and inference engine cannot learn, that is to say, they cannot adapt to changing environments. Learning must be incorporated with the addition of cased-based reasoning and/or neural networks. Expert systems lack sensory experience, thus it is dependent on symbolic input. Expert system are not good a recognizing when no answer exists. It is interesting to note that every disadvantage of an expert system is rooted in the fact that it is not a human, therefore it cannot exercise common sense, invoke creativity, sense it’s environment or say “I don’t know”.
Interestingly, it is the advantages and disadvantages of expert systems that determines the fields in which one can be built. Any area that is cut of from everyday common sense and social intercourse could support an expert system. But even when a field has been selected that can support an expert system, extracting the expert knowledge from the expert can be a daunting task. Many Knowledge Engineers agree that while an expert in a particular field may be able to give expert advice, he/she usually has difficulty articulating the principles underlying their expertise.
The trend with expert systems is that they are increasingly being used for tasks that involve social interaction. Some have even suggested that they (ES) can take the place of visiting a physician in order to gain a diagnosis for some malady. For instance, easydiagnosis.com (by matheMEDics) employs the use of expert systems to offer on-line diagnosis of many illnesses by listing an estimated order of likelihood and estimating probabilities of diseases or conditions based on its internal logic and your answers to the questions.
Upon entrance to this website, a disclaimer is displayed that advises the user not to substitute this diagnosis for a consultation with a physician. Once you have electronically accepted this disclaimer, a page is displayed which has numerous questions that are formatted in such a way as to allow the expert system to draw a conclusion about your stated condition. Once all of the questions are answered, you click on the evaluate button and your diagnosis is displayed.
This situation is a prime example of why an expert system should be limited to those fields that do not require social intercourse. If a user decides to lie about his/her condition then his/her diagnosis will be inaccurate. Thus, in these types of settings, an expert system can be manipulated to give a diagnosis that is untrue and inaccurate. I am reminded of the magazine questionnaires that allowed you to rate your love life, or some other aspect of your life. Many times the user completed the questionnaire in such a way as to get a favorable rating rather than a true rating. In like manner, the aforementioned expert system and those like it would be of little value to the user and hardly worth the time and effort it takes to build one.
In conclusion, it is my opinion
that expert systems can be of great assistance to mankind, but they should
be limited to those areas that do not require social interaction, since the validity of the conclusion an
expert system gives is only as
accurate as the response to the questions it asks.
References:
Applications: Graduate school Admissions Screening, Acquired Intelligence Inc.
Applications: The Employment Coach, Acquired Intelligence Inc.
Knowledge Based Expert System Services at Acquired Intelligence, Acquired Intelligence Inc.
Introduction
to Artificial Intelligence and Expert Systems,
Carol E. Brown, Daniel E. O’Leary --
-
accouting.Rutger.edu/raw/es…
Think About it: Artificial Intelligence and Expert Systems, Steven Sweet
Expert Systems, Prof. Ben Avi
– www.ee.cooper.edu/courses/course_pages/past/EE/459/expert/
- www.inf.ufsc.br/~mauro/ine6102/leituras/socrates.html
On-line diagnosis of Alcoholism, matheMEDics, Inc.
- easydiagnosis.com