You are on page 1of 6

Expert Systems

BEFORE YOU START

• Do you know that computers can be used to aid decision making?


• Do you know what is meant by a database?

An expert system operates by prompting the user to


Introduction enter certain data using the user interface, referring to
the knowledge base and using the inference engine to aid
You spend your life developing your knowledge,
the decision-making process it is designed to simulate.
thinking and making decisions. Decisions are
fundamental parts of life. It is also possible to Expert systems are versatile and can be used in a range
develop systems to aid this process and they are called of different applications. Expert systems can be used as
expert systems. a diagnostic tool, in financial planning and risk analysis.
Expert systems can even be used for a more fun purpose,
An expert system is a computerised system that attempts
such as a challenging chess opponent. They are also a
to reproduce the decision-making process of an expert
form of artificial intelligence which is a computer system
human being. The person who designs and develops an
designed to simulate human intelligence.
expert system is called an expert system engineer. They are
designed to try and replicate the judgement of a human
that has expert knowledge in a certain field. By doing this KEY WORDS
they can be used to replace or assist a human expert.
artificial intelligence: a computer system
designed to simulate human intelligence

7.1 Components of an
expert system
User interface
The user interface is the way that a user interacts with
the expert system. This could include using a keyboard
to enter criteria into text query boxes, or choosing
options by pressing offered choices on a touch screen.
Figure 7.1: Thinking about decisions. The user interface will guide the user about what data
they need to input into the expert system and will then
An expert system consists of several components. display any output from the expert system.
These components include: Without the presence of a user interface, a user would
• a user interface need to know how to program each of the interactions
they want to make with the expert system. The quality
• a knowledge base of the design of a user interface is very important.
• an inference engine
• a knowledge base editor
• an explanation system
• method of output, e.g. screen.
A simple example of a rule could be:
IF a > b AND a > c THEN highest = a

Figure 7.2: Touch screen interface.

Knowledge base Figure 7.3: Experts pooling knowledge.


The knowledge base is a database of related information
about a particular subject. It allows the storage and The knowledge base should be a dynamic resource,
retrieval of the knowledge required for an expert system therefore the expert system needs to have a knowledge
to operate. base editor. The knowledge base editor allows the
When an expert system is developed, several experts will knowledge base to be edited and updated when
be interviewed and asked to contribute the knowledge necessary.
they have of a given field. This knowledge is then used
to build a database that is the knowledge base for the
expert system. The developers will want two types of KEY WORDS
knowledge from the experts, factual knowledge and knowledge base editor: a component of an
heuristic knowledge. Factual knowledge is knowledge expert system that is used to amend or update
that is definitive and widely shared amongst experts the knowledge base
in the field. Heuristic knowledge is knowledge that is
acquired through personal experiences and built on
reasoning.
Part of the knowledge base is the rules base. The rules
Question
base is a set of rules that will be used to produce an 2 What could happen if the expert system did not
output or decision by the expert system. These rules are have a knowledge base editor?
used by the inference engine as a base for reasoning, to
obtain a solution to a problem or a decision. Each rule
will contain two parts, the IF and the THEN. A rule can Inference engine
also have multiple IF parts that will be joined together The inference engine is the part of the expert system that
by Boolean operators including AND and OR. makes judgements and reasoning using the knowledge
base and user responses. It is designed to produce
KEY WORDS reasoning based on the rules and the knowledge base.
It will ask the user questions and, based on their answer,
knowledge base: a component of an expert it will follow a line of logic. This may then lead to
system that stores the knowledge provided further questions and, eventually, to a final result.
by experts
The inference engine is mostly a problem-solving tool. It
rules base: a part of the knowledge base that organises and controls the steps to providing the desired
contains all the rules to be analysed by the output. There are two main methods that an inference
expert system engine can use to simulate reasoning, these are backward
chaining and forward chaining.
Backward chaining is based on goal driven reasoning,
i.e. is dependent on a finding a desired goal. This type
Explanation system
of chaining is used when the possible outcomes are The conclusion or decision the expert system provides
limited and definitive in nature. In backward chaining, may not always be an obvious choice to a user. The
the system tries to take a goal and repeatedly split it into user may want to gain an understanding of how the
sub-goals that are simpler to achieve. The nature of this conclusion or decision was determined. In order
type of system is that it moves backward from the goal to to allow this facility, some expert systems have an
be achieved, using the sub-goals to inform the next piece explanation system built into them. This will provide an
of data that is needed by the system, to reach a goal. explanation of the reasoning process and show how the
output given by the system was achieved.
A simple diagram that represents the process of
backward chaining would be:
KEY WORD
Sub-goals Rules Goal
explanation system: a component of an expert
Figure 7.4: Backward chaining. system that provides an explanation of how an
outcome was achieved
Forward chaining is based on data driven reasoning and
is dependent on the data that it is provided with. This
type of system is used when a problem is more open
ended, and the outcome is not necessarily definitive in Method of output
nature. The system will take data input by the user and The output method is the method the user will use to
move forward from rule to rule to suggest a possible view any results produced by the expert system. This
outcome. It will take the data input by a user, then move will often be in the form of a display screen that will
from rule to rule to find one where the clause for the allow them to see the results on screen, or may include
data input is true. It will then ask the user for more data an output, such as a printer, that allows the results to be
and repeat this process until it can suggest an outcome. printed and viewed.

KEY WORDS
7.2 Are expert systems
backward chaining: breaking a goal down
into sub-goals that allow the system to work
backward from the goal
useful?
There are several advantages that can be gained from
forward chaining: a system that moves using an expert system.
forward from rule to rule until it reaches a
Expert systems:
possible outcome
• can provide answers to questions that are outside
goal driven: a system that is dependent on a the knowledge that you currently have
finding a desired goal • can aid professional people by prompting them and
data driven: a system dependent on the data guiding them to look at areas of knowledge they
that it is provided with may not have considered or remembered
• are consistent in the responses they produce as they
are arrived at in a logical way
A simple diagram that represents the process of forward • can be used at any time, so you do not need
chaining would be: to contact another person who may have the
knowledge at an unsuitable time
Data Rules Outcome
• can sometimes arrive at a solution to a problem
Figure 7.5: Forward chaining. quicker than a human would.
There are clear advantages to using an expert system,
however there are also disadvantages to using
them as well.
Expert systems: and then provides possible conditions that match the
• do not have the intuition that humans have. This symptoms. It lists the possible conditions in order,
means that their response can only be a logical one starting with the one it finds to be the closest match.
and may not be useful.
• are only as good as the rules and data they are
provided with. If there are errors in the data or
rules, then this will produce incorrect results.
• are expensive to create. Many experts need to be
consulted and a high level of skill is required to
build the component parts.
• cannot adapt a great deal to their environment and
may require the knowledge base to be edited in
order to do this.

Questions Figure 7.6: Medical diagnosis interface.


3 Think of a scenario in which an expert system
would be an advantage.
4 Think of a scenario where an expert system would PRACTICAL ACTIVITY 7.01
not be very suitable.
Do you think a medical expert system will be
a data driven or goal driven expert system?
REFLECTION Discuss with a partner your thoughts about which
it could be and why.
Do you think you have ever used an expert system?

Question
7.3 How are expert 5 What could be the benefits and the disadvantages
of people using online self-diagnosis systems?
systems used?
Expert systems are used by many individuals and Car engine fault diagnosis
organisations for a variety of different reasons. Most modern cars have an abundance of technology
These include: built into them. Car mechanics will often be very
• medical diagnosis knowledgeable about the mechanical parts of the car,
• car engine fault diagnosis but they may not have as much knowledge about the
modern technology.
• a digital opponent in games such as chess
• providing financial planning and investment advice The cars often have a system of symbols on their
dashboard that light up when an issue is detected in the
• providing insurance planning advice
car, such as a problem with the engine. The driver of the
• plant and animal identification car sees the engine symbol light up and this shows that
• planning and scheduling routes for delivery vehicles the car may need to be taken to a garage.
• mineral prospecting. When the driver takes the car to a garage, the mechanic
plugs the expert system into the car engine. The expert
system then interrogates the engine management system
Medical symptoms to discover all the information available about the fault,
The organisation WebMD have a website that offers a so that it can provide a diagnosis. The expert system
medical expert system called ‘symptom checker’. It asks provides a list of possible diagnoses to the mechanic,
the user a series of questions about the user’s symptoms who can then use their knowledge to fix the car.
PRACTICAL ACTIVITY 7.02

Find out what Deep Blue was developed to do


after its chess victory.

PRACTICAL ACTIVITY 7.03

Find out about another chess playing system


called AlphaZero.

Figure 7.7: Car fault warnings. Financial planning


Expert systems can be a very useful tool in financial
planning. They can help individuals on a small scale or
Chess player companies on a large scale. They can create a financial
plan or assess what financial actions need to be taken to
One of the most famous examples of an expert system achieve a desired outcome.
used as a chess player is a computer called Deep Blue.
Deep Blue was developed by the organisation IBM. They can help people to:
On 11 May 1997, Deep Blue beat the world champion • manage their debt
chess player Garry Kasparov. The victory received press
• organise their investments
coverage around the world, as people were amazed
at the abilities of the computer. The success of Deep • reduce their taxes
Blue inspired computer scientists around the world to
• plan for retirement.
develop the abilities of expert systems much further.
A user enters information about their financial situation
People were quite sceptical about the victory of Deep
and personal circumstances. The expert system will then
Blue. They could not believe that a computer could play
produce a financial plan, or a list of possible outcomes,
with the capabilities that it showed. It was one of the first
based on this information.
realisations that computers had great capabilities when it
came to simulating thinking. This scared some people.
Deep Blue is now stored in the Smithsonian Museum in
Washington DC.

Figure 7.9: Financial planning.

Figure 7.8: Computer chess master.


7 Expert systems

Insurance planning
Many insurance company websites use an online expert
system so that users can quickly see if the company has
a policy that would be suitable for them. The system
will ask a user for their circumstances and it will then
provide a result to say whether the insurance company
would be willing to give the user an insurance policy.
If the expert system is able to match their data to a

the policy, such as how much it will cost and what


benefits it will provide.
Figure 7.10: Identifying plants.

possible policy, it will inform the user with details about

Planning a route for a delivery


Plant identification company
A website called Botanical keys allows users to enter a
range of characteristics about a plant. It mostly covers Delivery companies can use expert systems to find out
plants in the UK and Ireland area. the best route for a delivery driver for all the deliveries
the driver needs to make on a given day.
The website takes the data input about the plant and
analyses it against its rules and knowledge base. It then Each of the deliveries the driver needs to make is put
tells the user what kind of plant it could be. into the system. The system then outputs a route that
will be the most efficient route for the driver to follow,
A user could find out what kind of plant is growing in in terms of time or fuel used, etc. to make each delivery.
their garden, or try to discover the type of plant they It may request further information, such as the type of
have seen when on a walk. delivery vehicle and how long each delivery will take.
The system may make further recommendations about
how many drivers are needed to deliver all the parcels
that day and what kind of vehicle each driver will need
to take for the parcels they need to deliver.
It is likely that an expert system could work out this type
of outcome quicker than a human, due to the amount
of data it is likely to handle, especially in large delivery
companies.

You might also like