Expert Systems

• An expert system, also known as decision support systems, is software that contains the knowledge and analytical skills of one or more human experts. It applies reasoning methodologies to knowledge in a specific domain to render advice or recommendations, much like a human expert. (Aronson & Turban 2001, p811)
• Expert systems are also known as knowledge based systems or symbolic systems because “they transform symbols representing things in the real world into other symbols according to explicit rules” (Aikenhead M, 1996 at 33).
• Expert systems allow organisations to ‘capture’ and use the wisdom of experts and specialists in a particular field

• An expert system is an example of an intelligent system. Intelligent systems produce output similar to that expected from a human.

• An expert system is a piece of software that utilises facts and rules.
• These facts and rules are stored in a knowledge based that is created by a human expert.
• The human expert must encode this information using the correct syntax recognised by the software to create the knowledge base.

In business, there are three main areas where decisions are critical:

1. Maximising income
2. Minimising costs
3. Minimising risks

Two types of business decisions

• Short term – operational, day-to-day
• Long term – strategic decisions

Types of problems management can use DSS with include:

1. Descriptive – can use SQL. E.g. how many…
2. Investigative – spreadsheet analysis and WHAT IF processing
3. Explanation – data mining to work out patterns E.g. Why did…
4. Predictive
5. Prescriptive – E.g. what do we do for …

Expert Systems use a combination of software including models, analytical tools, databases, and automated processes. The solution can be enterprise-wide requiring huge amounts of storage and processing capabilities or they can be simple, PC-desktop systems. Regardless of scope and size, they have three main functions:

1. Data management – data integrity, structure, collecting and organising. Social issues such as security and privacy must be addressed.
2. Model management – a suitable model needs to be designed, built and tested.
3. Dialogue management – design of an appropriate interface that is human-centred, intuitive and easy to use.

• Increased productivity and output - expert systems can work faster than humans and can increase productivity, efficiency and save valuable resources.
• Permanent capturing of scarce expertise - expertise in particular areas are captured in the system permanently and making it particularly useful for retaining expertise that is scarce and uncommon.
• Creates knowledge accessibility - common or more routine knowledge captured in expert systems make the knowledge accessible by others using the system allowing users to receive relevant or useful advice.
• Can solve complex problems - complex problems and solutions may be solved by capturing a broad scope of knowledge from more than one source or individual.

(Aronson & Turban 2001, p420)

There are various issues that create problems or limit the widespread use of expert systems.

• Expert knowledge may not always be easily obtained or available when creating the system.
• Difficulties may be faced in extracting expert knowledge from humans.
• There may be various approaches to assessing situations by different experts yet all correct.
• There may be a lot of biases in the knowledge transfer process.

(Aronson & Turban 2001, p420)

• Process used by the inference engine to arrive at a goal.
• Starts at the first rule in the knowledge base and moves forward selectively eliminating irrelevant rules as the user enter current conditions until it arrives at an appropriate solution.
• Forward Chaining is best suited for systems where there are a large number of possible solutions.

• Using this approach, the inference engine assumes the most common goal.
• It then works backwards through the relevant rules for that goal, trying to establish that the necessary conditions exist. Excellent when there are few solutions.
• As soon as a necessary condition is not present, it discards that goal and starts again assuming the next most popular one is true.

There are a variety of factors which contribute to a good expert system.

• Excellent facts and rules - complete, consistent, relevant to the problem being solved
• Interface - Easy to use and well designed; unambiguous questions; Non tedious questions; Explanation facility
• Software engine - The logic must be tight and efficient; Has it been rigorously tested; Expert systems with many rules can be very rigorous to test, due to the many rules and combination of input data. (Often software is used to test the data).

1. Aronson J & Turban E, 2001, Decision Support Systems and Intelligent Systems, 6th ed., Prentice Hall, New Jersey.
2. Aikenhead M, “The uses and abuses of neural networks in law”, (1996) 12 Santa Clara Computer and High Technology Law Journal 31.