Applications of Artificial

Intelligence UNIT 1 EXPERT SYSTEMS

Structure Page Nos.

1.0 Introduction 5
1.1 Objectives 5
1.2 An Introduction to Expert Systems 6
1.3 Concept of Planning, Representing and using Domain Knowledge 7
1.4 Knowledge Representation Schemes 7
1.4.1 Semantic Networks
1.4.2 Frames
1.4.3 Proposition and Predicate Logic
1.4.4 Rule Based Systems
1.4.4.1 Forward Chaining Systems
1.4.4.2 Backward Chaining Systems
1.4.4.3 Probability Certainty Factors in Rule Based Systems
1.5 Examples of Expert Systems: MYCIN, COMPASS 24
1.6 Expert System Building Tools 25
1.6.1 Expert System Shells
1.6.1.1 Knowledge Base
1.6.1.2 Knowledge Acquisition Subsystem (Example: COMPASS)
1.6.1.3 Inference Engine
1.6.1.4 Explanation Sub-system (Example: MYCIN)
1.6.1.5 User Interface
1.6.1.6 An Expert System shell: EMYCIN
1.7 Some Application Areas of Expert Systems 32
1.8 Summary 32
1.9 Solutions/Answers 33
1.10 Further Readings 35

1.0 INTRODUCTION

Computer Science is the study of how to create models that can be represented in and
executed by some computing equipment. A number of new types of problems are
being solved almost everyday by developing relevant models of the domains of the
problems under consideration. One of the major problems which humanity encounters
is in respect of scarcity of human experts to handle problems from various domains of
human experience relating to domains including domains those of health, education,
economic welfare, natural resources and the environment. In this respect, the task for
a computer scientist is to create, in addition to a model of the problem domain, a
model of an expert of the domain as problem solver who is highly skilled in solving
problems from the domain under consideration. The field of Expert Systems is
concerned with creating such models. The task includes activities related to eliciting
information from the experts in the domain in respect of how they solve the problems,
and activities related to codifying that information generally in the form of rules. This
unit discusses such issues about the design and development of an expert system.

1.1 OBJECTIVES

After going through this unit, you should be able to:

 discuss the various knowledge representation scheme;
 tell about some of the well-known expert systems;
 enumerate and use of various tools for building expert systems;
 explain the various expert system shells, and
 discuss various application areas for expert systems.
5
 Expert Systems
1.2 AN INTRODUCTION TO EXPERT SYSTEMS

First of all we must understand that an expert system is nothing but a computer
program or a set of computer programs which contains the knowledge and some
inference capability of an expert, most generally a human expert, in a particular
domain. As expert system is supposed to contain the capability to lead to some
conclusion based on the inputs provided, information it already contains and its
processing capability, an expert system belongs to the branch of Computer Science
called Artificial Intelligence.

Mere possessing an algorithm for solving a problem is not sufficient for a program to
be termed an expert system, it must also possess knowledge i.e., if there is an expert
system for a particular domain or area and if it is fed with a number of questions
regarding that domain then sooner or later we can expect that these questions will be
answered. So we can say that the knowledge contained by an expert system must
contribute towards solving the problems for which it has been designed.

Also knowledge in a expert system must be regarding a specific domain. As a human

being cannot be an expert in every area of life, similarly, an expert system which tries
to simulate the capabilities of an expert also works in a particular domain. Otherwise
it may be require to possess potentially infinite amount of knowledge and processing
that knowledge in finite amount of time is an impossible task.

Taking into consideration all the points which have been discussed above, let us try to
give one of the many possible definitions of an Expert System:

An Expert System is a computer program that possesses or represents knowledge in a

particular domain, has the capability of processing/ manipulating or reasoning with
this knowledge with a view to solving a problem, giving some achieving or to achieve
some specific goal.

An expert system may or may not provide the complete expertise or functionality of a
human expert but it must be able to assist a human expert in fast decision making. The
program might interact with a human expert or with a customer directly.

Let us discuss some of the basic properties of an expert system:

 It tries to simulate human reasoning capability about a specific domain rather

than the domain itself. This feature separates expert systems from some other
familiar programs that use mathematical modeling or computer animation. In an
expert system the focus is to emulate an expert‟s knowledge and problem solving
capabilities and if possible, at a faster rate than a human expert.
 It perform reasoning over the acquired knowledge, rather than merely performing
some calculations or performing data retrieval.
 It can solve problems by using heuristic or approximate models which, unlike
other algorithmic solutions are not guaranteed to succeed.

AI programs that achieve expert-level competence in solving problems in different

domains are more called knowledge based systems. A knowledge-based system is
any system which performs a job or task by applying rules of thumb to a symbolic
representation of knowledge, instead of employing mostly algorithmic or statistical
methods. Often the term expert systems is reserved for programs whose knowledge
base contains the knowledge used by human experts, in contrast to knowledge
gathered from textbooks or non-experts. But more often than not, the two terms,
expert systems and knowledge-based systems are taken us synonyms. Together
6
Applications of Artificial they represent the most widespread type of AI application. The area of human
Intelligence intellectual endeavour to be captured in an expert system is sometimes called the task
domain. Task refers to some goal-oriented, problem-solving activity. Domain refers
to the area within which the task is being performed. Some of the typical tasks are
diagnosis, planning, scheduling, configuration and design. For example, a program
capable of conversing about the weather would be a knowledge-based system, even if
that program did not have any expertise in meteorology, but an expert system must be
able to perform weather forecasting.

Building a expert system is known as knowledge engineering and its practitioners

are called knowledge engineers. It is the job of the knowledge engineer to ensure to
make sure that the computer has all the knowledge needed to solve a problem. The
knowledge engineer must choose one or more forms in which to represent the
required knowledge i.e., s/he must choose one or more knowledge representation
schemes (A number of knowledge representing schemes like semantic nets, frames,
predicate logic and rule based systems have been discussed above. We would be
sticking to rule based representation scheme for our discussion on expert systems).
S/he must also ensure that the computer can use the knowledge efficiently by selecting
from a handful of reasoning methods.

1.3 CONCEPT OF PLANNING, REPRESENTING

AND USING DOMAIN KNOWLEDGE

From our everyday experience, we know that in order to solve difficult problems, we
need to do some sort of planning. Informally, we can say that Planning is the process
that exploits the structure of the problem under consideration for designing a sequence
of actions in order to solve the problem under consideration.

The knowledge of nature and structure of the problem domain is essential for planning
a solution of the problem under consideration. For the purpose of planning, the
problem environments are divided into two categories, viz., classical planning
environments and non-classical planning environments. The classical planning
environments/domains are fully observable, deterministic, finite, static and discrete.
On the other hand, non-classical planning environments may be only partially
observable and/or stochastic. In this unit, we discuss planning only for classical
environments.

1.4 KNOWLEDGE REPRESENTATION SCHEMES

One of the underlying assumptions in Artificial Intelligence is that intelligent

behaviour can be achieved through the manipulation of symbol structures
(representing bits of knowledge). One of the main issues in AI is to find appropriate
representation of problem elements and available actions as symbol structures so that
the representation can be used to intelligently solve problems. In AI, an important
criteria about knowledge representation schemes or languages is that they should
support inference. For intelligent action, the inferencing capability is essential in
view of the fact that we can‟t represent explicitly everything that the system might
ever need to know–somethings have to be left implicit, to be inferred/deduced by
the system as and when needed in problem solving.

In general, a good knowledge representation scheme should have the following

features:

7
 It should allow us to express the knowledge we wish to represent in the language. Expert Systems
For example, the mathematical statement: Every symmetric and transitive
relation on a domain, need not be reflexive is not expressible in First Order
Logic.
 It should allow new knowledge to be inferred from a basic set of facts, as
discussed above.
 It should have well-defined syntax and semantics.

Some popular knowledge representation schemes are:

 Semantic networks,
 Frames,
 First order logic, and,
 Rule-based systems.

As semantic networks, frames and predicate logic have been discussed in previous
blocks so we will discuss these briefly here. We will discuss the rule-based systems
in detail.

1.4.1 Semantic Networks

Semantic Network representations provide a structured knowledge representation.

In such a network, parts of knowledge are clustered into semantic groups. In semantic
networks, the concepts and entities/objects of the problem domain are represented by
nodes and relationships between these entities are shown by arrows, generally, by
directed arrows. In view of the fact that semantic network representation is a
pictorial depiction of objects, their attributes and the relationships that exist between
these objects and other entities. A semantic net is just a graph, where the nodes in the
graph represent concepts, and the arcs are labeled and represent binary relationships
between concepts. These networks provide a more natural way, as compared to other
representation schemes, for mapping to and from a natural language.

For example, the fact (a piece of knowledge): Mohan struck Nita in the garden with
a sharp knife last week, is represented by the semantic network shown in Figure 4.1.

struck

past of
time agent
last week strike Mohan

instrument
place
object knife

garden

Nita
property of

sharp

Figure 1.1 Semantic Network of the Fact in Bold

8
Applications of Artificial
Intelligence The two most important relations between concepts are (i) subclass relation between a
class and its superclass, and (ii) instance relation between an object and its class.
Other relations may be has-part, color etc. As mentioned earlier, relations are
indicated by labeled arcs.

As information in semantic networks is clustered together through relational

links, the knowledge required for the performance of some task is generally available
within short spatial span of the semantic network. This type of knowledge
organisation in some way, resembles the way knowledge is stored and retrieved by
human beings.

Subclass and instance relations allow us to use inheritance to infer new

facts/relations from the explicitly represented ones. We have already mentioned that
the graphical portrayal of knowledge in semantic networks, being visual, is easier
than other representation schemes for the human beings to comprehend. This fact
helps the human beings to guide the expert system, whenever required. This is perhaps
the reason for the popularity of semantic networks.

Exercise 1: Draw a semantic network for the following English statement:

Mohan struck Nita and Nita’s mother struck Mohan.

1.4.2 Frames

Frames are a variant of semantic networks that are one of the popular ways of
representing non-procedural knowledge in an expert system. In a frame, all the
information relevant to a particular concept is stored in a single complex entity, called
a frame. Frames look like the data structure, record. Frames support inheritance. They
are often used to capture knowledge about typical objects or events, such as a car, or
even a mathematical object like rectangle. As mentioned earlier, a frame is a
structured object and different names like Schema, Script, Prototype, and even Object
are used in stead of frame, in computer science literature.

We may represent some knowledge about a lion in frames as follows:

Mammal :
Subclass : Animal
warm_blooded : yes

Lion :
subclass : Mammal
eating-habbit : carnivorous
size : medium

Raja :
instance : Lion
colour : dull-Yellow
owner : Amar Circus

Sheru :
instance : Lion
size : small

9
A particular frame (such as Lion) has a number of attributes or slots such as eating- Expert Systems
habit and size. Each of these slots may be filled with particular values, such as the
eating-habit for lion may be filled up as carnivorous.

Sometimes a slot contains additional information such as how to apply or use the slot
values. Typically, a slot contains information such as (attribute, value) pairs, default
values, conditions for filling a slot, pointers to other related frames, and also
procedures that are activated when needed for different purposes.

In the case of frame representation of knowledge, inheritance is simple if an object

has a single parent class, and if each slot takes a single value. For example, if a
mammal is warm blooded then automatically a lion being a mammal will also be
warm blooded.

But in case of multiple inheritance i.e., in case of an object having more than one
parent class, we have to decide which parent to inherit from. For example, a lion may
inherit from “wild animals” or “circus animals”. In general, both the slots and slot
values may themselves be frames and so on.

Frame systems are pretty complex and sophisticated knowledge representation

tools. This representation has become so popular that special high level frame based
representation languages have been developed. Most of these languages use LISP as
the host language. It is also possible to represent frame-like structures using object
oriented programming languages, extensions to the programming language LISP.

Exercise 2: Define a frame for the entity date which consists of day, month and year.
each of which is a number with restrictions which are well-known. Also a procedure
named compute-day-of-week is already defined.

1.4.3 Proposition and Predicate Logic

Symbolic logic may be thought of as a formal language, which incorporates a precise

system for reasoning and deduction. Propositional logic and predicate logic are two
well-known forms of symbolic logic. Propositional logic deals with propositions or
statements.

A proposition or a statement is a sentence, which can be assigned a truth-value true or

false. For example, the statement:

The sun rises in the west, has a truth value false.

On the other hand, none of the following sentences can be assigned a truth-value, and
hence none of these, is a statement or proposition:

(i) Ram must exercise regularly. (Imperative sentence)

(ii) Who was the first Prime Minister of India? (Interrogative sentence)
(iii) Please, give me that book. (Imperative sentence)
(iv) Hurry! We have won the trophy. (Exclamatory sentence).

Generally, the following steps are followed for solving problems using
propositional logic, where the problems are expressed in English and are such that
these can be solved using propositional logic (PL):

a) First problem statements in English are translated to formulae of Propositional

logic

10
Applications of Artificial b) Next, some of the rules of inference in PL including the ones mentioned
Intelligence below, are used to solve the problem, if, at all, the problem under
consideration is solvable by PL.

Some of the rules of Inference of Propositional Logic:

(i) Modus Ponens P

PQ
Q
(The above notation is interpreted as: if we are given the two formulae P and
PQ of the propositional logic, then conclude the formula Q. In other words, if
both the formulae P and PQ are assumed to be true, then by modus ponens, we
can assume the statement Q also as true.)

(ii) Chain Rule

PQ
QR
PR
(iii) Transposition
P Q
 Q P
Example

Given the following three statements:

(i) Matter always existed
(ii) If there is God, then God created the universe
(iii) If God created the universe, then matter did not always exist.

To show the truth of the statement: There is no God.

Solution:
Let us denote the atomic statements in the argument given above as follows:
M: Matter always existed
TG: There is God
GU: God created the universe.
Then the given statements in English, become respectively the formulae of PL:
(i) M
(ii) TGGU
(iii) GU  M
(iv) To show  TG
Applying transposition to (iii) we get
(v) M GU
using (i) and (v) and applying Modus Ponens, we get
(vi) GU
Again, applying transposition to (ii) we get
(vii) GU TG
Applying Modus Ponens to (vi) and (vii) we get
(viii) TG
The formula (viii) is the same as formula (iv) which was required to be proved.

Exercise 3: Using prepositional logic, show that, if the following statements are
assumed to be true:
(i) There is a moral law.
(ii) If there is a moral law, then someone gave it.
11
(iii) If someone gave the moral law, then there is God. Expert Systems
then the following statement is also true:
(iv) There is God.

The trouble with propositional logic is that it is unable to describe properties of

objects and also it lacks the structure to express relations that exist among two or more
entities. Further, propositional logic does not allow us to make generalised statements
about classes of similar objects. However, in many situations, the explicit knowledge
of relations between objects and generalised statements about classes of similar
objects, are essentially required to solve problems. Thus, propositional logic has
serious limitations while reasoning for solving real-world problems. For example, let
us look at the following statements:

(i) All children more than 12 years old must exercise regularly.
(ii) Ram is more than 12 years old.

Now „these statements should be sufficient enough to allow us to conclude: Ram must
exercise regularly. However, in propositional logic, each of the above statement is
indecomposable and may be respectively denoted by P and Q. Further, whatever is
said inside each statement is presumed to be not visible. Therefore, if we use the
language of propositional logic, we are just given two symbols, viz., P and Q,
representing respectively the two given statements. However, from just two
propositional formulae P and Q, it is not possible to conclude the above mentioned
statement viz., Ram must exercise regularly. To draw the desired conclusion with a
valid inference rule, it would be necessary to use some other language, including some
extension of propositional logic.

Predicate Logic, and more specifically, First Order Predicate Logic (FOPL) is an
extension of propositional logic, which was developed to extend the expressiveness of
propositional logic. In addition to just propositions of propositional logic, the
predicate logic uses predicates, functions, and variables together with variable
quantifiers (Universal and Existential quantifiers) to express knowledge.

We already have defined the structure of formulae of FOPL and also have explained
the procedure for finding the meaning of formulae in FOPL. Though, we have already
explained how to solve problems using FOPL, yet just for recalling the procedure for
solving problems using FOPL, we will consider below one example.

In this context, we may recall the inference rules of FOPL. The inference rules of PL
including Modus Ponens, Chain Rule and Rule of Transposition are valid in FOPL
also after suitable modifications by which formulae of PL are replaced by formulae of
FOPL.

In addition to these inference rules, the following four inference rules of FOPL, that
will be called Q1, Q2, Q3 and Q4, have no corresponding rules in PL. In the following F
denotes a predicate and x a variable/parameter:

~ (x) F ( x) (x) ~ F ( x)
Q1: and
(x) ~ F ( x) ~ (x) F ( x)

The first of the above rules under Q1, says:

From negation of there exists x F (x), we can infer for all x not of F (x)

12
Applications of Artificial ~ (x) F ( x) (x) ~ F ( x)
Intelligence Q2: and
(x) ~ F ( x) ~ (x) F ( x)

The first of the above rules under Q2, says:

From negation of for all x F (x), we can infer there exists x such that not of F (x)

(x) F ( x)
Q3: , where a is (any) arbitrary element of the domain of F
F (a)
The rule Q3 is called universal instantiation

 F (a), for arbitrary a

Q 3:
(x) F ( x)

The rule is called universal generalisation

(x) F ( x)
Q4: , where a is a particular (not arbitrary) constant.
F (a)

This rule is also called existential instantiation:

 F (a) for some a

Q 4:
(x) F ( x)
The rule is called existential generalisation

Steps for using Predicate Calculus as a Language for Representing

Knowledge

Step 1: Conceptualization: First of all, all the relevant entities and the relations that
exist between these entities are explicitly enumerated. Some of the implicit facts like
„a person dead once is dead forever‟ have to be explicated.

Step 2: Nomenclature and Translation: Giving appropriate names to the objects and
relations. And then translating the given sentences given in English to formulae in
FOPL.

Appropriate names are essential in order to guide a reasoning system based on FOPL.
It is well-established that no reasoning system is complete. In other words, a reasoning
system may need help in arriving at desired conclusion.

Step 3: Finding appropriate sequence of reasoning steps, involving selection of

appropriate rule and appropriate FOPL formulae to which the selected rule is to be
applied, to reach the conclusion.

While solving problems with FOPL, generally, the proof technique is proof by
contradiction. Under this technique, the negation of what is to be proved is also taken
as one of the assumptions. Then from the given assumptions alongwith the new
assumption, we derive a contradiction, i.e., using inference rules, we derive a
statement which is negation of either an assumption or is negation of some earlier
derived formula.

Next, we give an example to illustrate how FOPL can be used to solve problems
expressed in English and which are solvable by using FOPL. However, the proposed
solution does not use the above-mentioned method of contradiction.

13
Example Expert Systems

We are given the statements:

(i) No feeling of pain is publically observable

(ii) All chemical processes are publically observable
We are to prove that
(iii) No feeling of pain is a chemical process

Solution:

For translating the given statements (i), (ii) & (iii), let us use the notation:

F(x): x is an instance of feeling of pain

O(x): x is an entity that is publically observable
C(x): x is a chemical process.

Then (i), (ii) and (iii) in FOPL can be equivalently expressed as

(i) (x) ( F(x)  O(x) )
(ii) (x) ( C(x) O(x) )

To prove
(iii) (x) ( F(x)  C(x) )

From (i) using generalized instantiation, we get

(iv) F(a)  O(a) for any arbitrary a.

Similarly, from (ii), using generalized instantiation, we get

(v) C(b)  O(b) for any arbitrary b.

From (iv) using transposition rule, we get

(vi) O(a)  F(a) for any arbitrary a

As b is arbitrary in (v), therefore we can rewrite (v) as

(vii) C(a)  O(a) for any arbitrary a
From (vii) and (vi) and using chain rule, we get
(viii) C(a)  F(a) for any arbitrary a

But as a is arbitrary in (viii), by generalized quantification, we get

(ix) (x) ( C(x)  F(x))

But (ix) is the same as (iii), which was required to be proved.

Problems with FOPL as a system of knowledge representation and reasoning:

FOPL is not capable of easily representing some kinds of information including
information pieces involving

(i) properties of relations. For example, the mathematical statement:

Any relation which is symmetric & transitive may not be reflexive
is not expressible in FOPL.
(ii) linguistic variables like hot, tall, sweat.
For example: It is very cold today,
can not be appropriately expressed in FOPL.
(iii) different belief systems.
For example, s/he know that he thinks India will win the match, but I think India will
lose,
also, can not be appropriately expressed in FOPL.
14
Applications of Artificial Some shortcomings in predicate calculus
Intelligence
Now, after having studied predicate logic as a knowledge representation scheme, we
ask ourselves the inevitable question, whether it can be used to solve real world
problems and to what extent.

Before we try to answer the above question, let us review some of the properties of
logic reasoning systems including predicate calculus. We must remember that three
important properties of any logical reasoning system are soundness, completeness and
tractability. To be confident that an inferred conclusion is true we require soundness.
To be confident that inference will eventually produce any true conclusion, we require
completeness. To be confident that inference is feasible, we require tractability.

Now, as we have discussed above also, in predicate calculus resolution refutation as

an inference procedure is sound and complete. Because in case that the well formed
formula which is to be proved is not logically followed or entailed by the set of well
formed formulas which are to be used as premises, the resolution refutation
procedure might never terminate. Thus, resolution refutation is not a full decision
procedure. Also there is not other procedure that has this property or that is fully
decidable. So predicate calculus is semi-decidable (semi-decidable means that if the
set of input (well formed formulas) do not lead to the conclusion then the procedure
will never stop or terminate).

But the situation is worse than this, as even on problems for which resolution
refutation terminates, the procedure is NP-hard – as is any sound and complete
inference procedure for the first-order predicate calculus i.e., it may take
exponentially large time to reach a conclusion.

How to tackle these problems:

People who have done research in Artificial Intelligence have shown various ways:
First, they say that we should not insist on the property of soundness of inference
rules. Now what does it mean – basically it means that sometimes or occasionally our
rules might prove an “untrue formula”.

Second, they say that we should not insist on the property of completeness i.e., to
allow use of procedures that are not guaranteed to find proofs of true formulas.

Third, they also suggest that we could use a language that is less expressive that the
predicate calculus. For example, a language in which everything is expressed using
only Horn Clauses ( Horn clauses are those which have at most one positive literal).

1.4.4 Rule Based Systems

Rather than representing knowledge in a declarative and somewhat static way (as a set
of statements, each of which is true), rule-based systems represent knowledge in terms
of a set of rules each of which specifies the conclusion that could be reached or
derived under given conditions or in different situations. A rule-based system consists
of
(i) Rule base, which is a set of IF-THEN rules,
(ii) A bunch of facts, and
(iii) Some interpreter of the facts and rules which is a mechanism which decides
which rule to apply based on the set of available facts. The interpreter also
initiates the action suggested by the rule selected for application.

15
 Expert Systems
A Rule-base may be of the form:
R1: If A is an animal and A barks, than A is a dog
F1: Rocky is an animal
F2: Rocky Barks
The rule-interpreter, after scanning the above rule-base may conclude: Rocky is a dog.
After this interpretation, the rule-base becomes

R1: If A is an animal and A barks, then A is a dog

F1: Rocky is an animal
F2: Rocky Barks
F3: Rocky is a dog.
There are two broad kinds of rule-based systems:
forward chaining systems,
and backward chaining systems.

In a forward chaining system we start with the initial facts, and keep using the rules
to draw new intermediate conclusions (or take certain actions) given those facts. The
process terminates when the final conclusion is established. In a backward chaining
system, we start with some goal statements, which are intended to be established and
keep looking for rules that would allow us to conclude, setting new sub-goals in the
process of reaching the ultimate goal. In the next round, the subgoals become the new
goals to be established. The process terminates when in this process all the subgoals
are given fact. Forward chaining systems are primarily data-driven, while backward
chaining systems are goal-driven. We will discuss each in detail.

Next, we discuss in detail some of the issues involved in a rule-based system.

Advantages of Rule-base

A basic principle of rule-based system is that each rule is an independent piece of

knowledge. In an IF-THEN rule, the IF-part contains all the conditions for the
application of the rule under consideration. THEN-part tells the action to be taken by
the interpreter. The interpreter need not search any where else except within the rule
itself for the conditions required for application of the rule.

Another important consequence of the above-mentioned characteristic of a rule-based

system is that no rule can call upon any other and hence rules are ignorant and hence
independent, of each other. This gives a highly modular structure to the rule-based
systems. Because of the highly modular structure of the rule-base, the rule-based
system addition, deletion and modification of a rule can be done without any danger
side effects.

Disadvantages

The main problem with the rule-based systems is that when the rule-base grows and
becomes very large, then checking (i) whether a new rule intended to be added is
redundant, i.e., it is already covered by some of the earlier rules. Still worse, as the
rule- base grows, checking the consistency of the rule-base also becomes quite
difficult. By consistency, we mean there may be two rules having similar conditions,
the actions by the two rules conflict with each other.

Let us first define working memory, before we study forward and backward chaining
systems.

Working Memory: A working is a representation, in which

16
Applications of Artificial  lexically, there are application –specific symbols.
Intelligence
 structurally, assertions are list of application-specific
symbols.
 semantically, assertions denote facts.
 assertions can be added or removed from working memory.

1.4.4.1 Forward Chaining Systems

In a forward chaining system the facts in the system are represented in a working
memory which is continually updated, so on the basis of a rule which is currently
being applied, the number of facts may either increase or decrease. Rules in the
system represent possible actions to be taken when specified conditions hold on items
in the working memory–they are sometimes called condition-action or antecedent-
consequent rules. The conditions are usually patterns that must match items in the
working memory, while the actions usually involve adding or deleting items from the
working memory. So we can say that in forward chaining proceeds forward,
beginning with facts, chaining through rules, and finally establishing the goal.
Forward chaining systems usually represent rules in standard implicational form, with
an antecedent or condition part consisting of positive literals, and a consequent or
conclusion part consisting of a positive literal.

The interpreter controls the application of the rules, given the working memory, thus
controlling the system‟s activity. It is based on a cycle of activity sometimes known as
a recognize-act cycle. The system first checks to find all the rules whose condition
parts are satisfied i.e., the those rules which are applicable, given the current state of
working memory (A rule is applicable if each of the literals in its antecedent i.e., the
condition part can be unified with a corresponding fact using consistent substitutions.
This restricted form of unification is called pattern matching). It then selects one and
performs the actions in the action part of the rule which may involve addition or
deleting of facts. The actions will result in a new i.e., updated working memory, and
the cycle starts again (When more than one rule is applicable, then some sort of
external conflict resolution scheme is used to decide which rule will be applied. But
when there are a large numbers of rules and facts then the number of unifications that
must be tried becomes prohibitive or difficult). This cycle will be repeated until either
there is no rule which fires, or the required goal is reached.

Rule-based systems vary greatly in their details and syntax, let us take the
following example in which we use forward chaining :

Example

Let us assume that the working memory initially contains the following facts :

(day monday)
(at-home ram)
(does-not-like ram)

Let, the existing set of rules are:

R1 : IF (day monday)
THEN ADD to working memory the fact : (working-with ram)

R2 : IF (day monday)
THEN ADD to working memory the fact : (talking-to ram)
17
 R3 : IF (talking-to X) AND (working-with X) Expert Systems
THEN ADD to working memory the fact : (busy-at-work X)

R4 : IF (busy-at-work X) OR (at-office X)
THEN ADD to working memory the fact : (not-at-home X)

R5 : IF (not-at-home X)
THEN DELETE from working memory the fact : (happy X)

R6 : IF (working-with X)
THEN DELETE from working memory the fact : (does-not-like X)

Now to start the process of inference through forward chaining, the rule based
system will first search for all the rule/s whose antecedent part/s are satisfied by the
current set of facts in the working memory. For example, in this example, we can see
that the rules R1 and R2 are satisfied, so the system will chose one of them using its
conflict resolution strategies. Let the rule R1 is chosen. So (working-with ram) is
added to the working memory (after substituting “ram” in place of X). So working
memory now looks like:

(working-with ram)
(day monday)
(at-home ram)
(does-not-like ram)

Now this cycle begins again, the system looks for rules that are satisfied, it finds rule
R2 and R6. Let the system chooses rule R2. So now (taking-to ram) is added to
working memory. So now working memory contains following:

(talking-to ram)
(working-with ram)
(day monday)
(at-home ram)
(does-not-like ram)

Now in the next cycle, rule R3 fires, so now (busy-at-work ram) is added to working
memory, which now looks like:

(busy-at-work ram)
(talking-to ram)
(working-with ram)
(day monday)
(at-home ram)
(does-not-like ram)

Now antecedent parts of rules R4 and R6 are satisfied. Let rule R4 fires, so (not-at-
home, ram) is added to working memory which now looks like :

(not-at-home ram)
18
Applications of Artificial (busy-at-work ram)
Intelligence
(talking-to ram)
(working-with ram)
(day monday)
(at-home ram)
(does-not-like ram)
In the next cycle, rule R5 fires so (at-home ram) is removed from the working
memory :

(not-at-home ram)
(busy-at-work ram)
(talking-to ram)
(working-with ram)
(day monday)
(does-not-like ram)
The forward chining will continue like this. But we have to be sure of one thing, that
the ordering of the rules firing is important. A change in the ordering sequence of
rules firing may result in a different working memory.

Some of the conflict resolution strategies which are used to decide which rule to fire
are given below:

 Don‟t fire a rule twice on the same data.

 Fire rules on more recent working memory elements before older ones. This
allows the system to follow through a single chain of reasoning, rather than
keeping on drawing new conclusions from old data.
 Fire rules with more specific preconditions before ones with more general
preconditions. This allows us to deal with non-standard cases.

These strategies may help in getting reasonable behavior from a forward chaining
system, but the most important thing is how should we write the rules. They
should be carefully constructed, with the preconditions specifying as precisely as
possible when different rules should fire. Otherwise we will have little idea or control
of what will happen.

1.4.4.2 Backward Chaining Systems

In forward chining systems we have seen how rule-based systems are used to draw
new conclusions from existing data and then add these conclusions to a working
memory. The forward chaining approach is most useful when we know all the
initial facts, but we don‟t have much idea what the conclusion might be.

If we know what the conclusion would be, or have some specific hypothesis to test,
forward chaining systems may be inefficient. In forward chaining we keep on moving
ahead until no more rules apply or we have added our hypothesis to the working
memory. But in the process the system is likely to do a lot of additional and irrelevant
work, adding uninteresting or irrelevant conclusions to working memory. Let us say
that in the example discussed before, suppose we want to find out whether “ram is at
home”. We could repeatedly fire rules, updating the working memory, checking each
time whether (at-home ram) is found in the new working memory. But maybe we
19
had a whole batch of rules for drawing conclusions about what happens when I‟m Expert Systems
working, or what happens on Monday–we really don‟t care about this, so would rather
only have to draw the conclusions that are relevant to the goal.

This can be done by backward chaining from the goal state or on some hypothesized
state that we are interested in. This is essentially how Prolog works. Given a goal state
to try and prove, for example (at-home ram), the system will first check to see if the
goal matches the initial facts given. If it does, then that goal succeeds. If it doesn‟t the
system will look for rules whose conclusions i.e., actions match the goal. One such
rule will be chosen, and the system will then try to prove any facts in the
preconditions of the rule using the same procedure, setting these as new goals to
prove. We should note that a backward chaining system does not need to update
a working memory. Instead it needs to keep track of what goals it needs to prove its
main hypothesis. So we can say that in a backward chaining system, the reasoning
proceeds “backward”, beginning with the goal to be established, chaining
through rules, and finally anchoring in facts.

Although, in principle same set of rules can be used for both forward and backward
chaining. However, in backward chaining, in practice we may choose to write the
rules slightly differently. In backward chaining we are concerned with matching the
conclusion of a rule against some goal that we are trying to prove. So the „then or
consequent‟ part of the rule is usually not expressed as an action to take (e.g.,
add/delete), but as a state which will be true if the premises are true.

To learn more, let us take a different example in which we use backward chaining
(The system is used to identify an animal based on its properties stored in the working
memory):

Example

1. Let us assume that the working memory initially contains the following facts:

(has-hair raja) representing the fact “raja has hair”

(big-mouth raja) representing the fact “raja has a big mouth”
(long-pointed-teeth raja) representing the fact “raja has long pointed teeth”
(claws raja) representing the fact “raja has claws”

Let, the existing set of rules are:

1. IF (gives-milk X)
THEN (mammal X)

2. IF (has-hair X)
THEN (mammal X)

3. IF (mammal X) AND (eats-meat X)

THEN (carnivorous X)

4. IF (mammal X) AND (long-pointed-teeth X) AND (claws X)

THEN (carnivorous X)

5. IF (mammal X) AND (does-not-eat-meat X)

THEN (herbivorous X)

6. IF (carnivorous X) AND (dark-spots X)

20
Applications of Artificial THEN (cheetah, X)
Intelligence
7. IF (herbivorous X) AND (long-legs X) AND (long-neck X) AND (dark-spots X)
THEN (giraffe, X)
8. IF (carnivorous X) AND (big-mouth X)
THEN (lion, X)

9. IF (herbivorous X) AND (long-trunk X) AND (big-size X)

THEN (elephant, X)

10. IF (herbivorous, X) AND (white-color X) AND ((black-strips X)

THEN (zebra, X)

Now to start the process of inference through backward chaining, the rule
based system will first form a hypothesis and then it will use the antecedent –
consequent rules (previously called condition – action rules) to work backward
toward hypothesis supporting assertions or facts.

Let us take the initial hypothesis that “raja is a lion” and then reason
about whether this hypothesis is viable using backward chaining approach explained
below :

 The system searches a rule, which has the initial hypothesis in the consequent part
that someone i.e., raja is a lion, which it finds in rule 8.

 The system moves from consequent to antecedent part of rule 8 and it finds the
first condition i.e., the first part of antecedent which says that “raja must be a
carnivorous”.

 Next the system searches for a rule whose consequent part declares that someone
i.e., “raja is a carnivorous”, two rules are found i.e., rule 3 and rule 4. We assume
that the system tries rule 3 first.

 To satisfy the consequent part of rule 3 which now has become the system‟s new
hypothesis, the system moves to the first part of antecedent which says that X i.e.,
raja has to be mammal.

 So a new sub-goal is created in which the system has to check that “raja is a
mammal”. It does so by hypothesizing it and tries to find a rule having a
consequent that someone or X is a mammal. Again the system finds two rules,
rule 1 and rule 2. Let us assume that the system tries rule 1 first.

 In rule 1, the system now moves to the first antecedent part which says that X
i.e., raja must give milk for it to be a mammal. The system cannot tell this
because this hypothesis is neither supported by any of the rules and also it is not
found among the existing facts in the working memory. So the system abandons
rule 1 and try to use rule 2 to establish that “raja is a mammal”.

 In rule 2, it moves to the antecedent which says that X i.e., raja must have hair for
it to be a mammal. The system already knows this as it is one of the facts in
working memory. So the antecedent part of rule 2 is satisfied and so the
consequent that “raja is a mammal” is established.

 Now the system backtracks to the rule 3 whose first antecedent part is satisfied. In
second condition of antecedent if finds its new sub-goal and in turn forms a new
hypothesis that X i.e., raja eats meat.

21
 The system tries to find a supporting rule or an assertion in the working memory Expert Systems
which says that “raja eats meat” but it finds none. So the system abandons the rule
3 and try to use rule 4 to establish that “raja is carnivorous”.

 In rule 4, the first part of antecedent says that raja must be a mammal for it to be
carnivorous. The system already knows that “raja is a mammal” because it was
already established when trying to satisfy the antecedents in rule 3.

 The system now moves to second part of antecedent in rule 4 and finds a new
sub-goal in which the system must check that X i.e., raja has long-pointed-teeth
which now becomes the new hypothesis. This is already established as “ raja has
long-pointed-teeth” is one of the assertions of the working memory.

 In third part of antecedent in rule 4 the system‟s new hypothesis is that “raja has
claws”. This also is already established because it is also one the assertions in the
working memory.

 Now as all the parts of the antecedent in rule 4 are established so its consequent
i.e., “raja is carnivorous” is established.

 The system now backtracks to rule 8 where in the second part of the antecedent
says that X i.e., raja must have a big-mouth which now becomes the new
hypothesis. This is already established because the system has an assertion that
“raja has a big mouth”.

 Now as the whole antecedent of rule 8 is satisfied so the system concludes that
“raja is a lion”.

We have seen that the system was able to work backward through the antecedent –
consequent rules, using desired conclusions to decide that what assertions it should
look for and ultimately establishing the initial hypothesis.

How to choose the type of chaining among forward or backward chaining for a
given problem ?

Many of the rule based deduction systems can chain either forward or backward, but
which of these approaches is better for a given problem is the point of discussion.

First, let us learn some basic things about rules i.e., how a rule relates its input/s
(i.e., facts) to output/s (i.e., conclusion). Whenever in a rule, a particular set of facts
can lead to many conclusions, the rule is said to have a high degree of fan out, and a
strong candidate of backward chaining for its processing. On the other hand,
whenever the rules are such that a particular hypothesis can lead to many questions for
the hypothesis to be established, the rule is said to have a high degree of fan in, and a
high degree of fan in is a strong candidate of forward chaining.

To summarize, the following points should help in choosing the type of chaining for
reasoning purpose :

 If the set of facts, either we already have or we may establish, can lead to a large
number of conclusions or outputs , but the number of ways or input paths to reach
that particular conclusion in which we are interested is small, then the degree of
fan out is more than degree of fan in. In such case, backward chaining is the
preferred choice.

22
Applications of Artificial  But, if the number of ways or input paths to reach the particular conclusion in
Intelligence which we are interested is large, but the number of conclusions that we can reach
using the facts through that rule is small, then the degree of fan in is more than
the degree of fan out. In such case, forward chaining is the preferred choice.

 For case where the degree of fan out and fan in are approximately same, then
in case if not many facts are available and the problem is check if one of the
many possible conclusions is true, backward chaining is the preferred choice.

1.4.4.3 Probability Certainty Factors in Rule Based System

Rule based systems usually work in domains where conclusions are rarely certain,
even when we are careful enough to try and include everything we can think of in the
antecedent or condition parts of rules.

Sources of Uncertainty

Two important sources of uncertainty in rule based systems are:

 The theory of the domain may be vague or incomplete so the methods to generate
exact or accurate knowledge are not known.
 Case data may be imprecise or unreliable and evidence may be missing or in
conflict.

So even though methods to generate exact knowledge are known but they are
impractical due to lack or data, imprecision or data or problems related to data
collection.

So rule based deduction system developers often build some sort of certainty or
probability computing procedure on and above the normal condition-action format
of rules. Certainty computing procedures attach a probability between 0 and 1 with
each assertion or fact. Each probability reflects how certain an assertion is, whereas
certainty factor of 0 indicates that the assertion is definitely false and certainty factor
of 1 indicates that the assertion is definitely true.

Example 1: In the example discussed above the assertion (ram at-home) may have a
certainty factor, say 0.7 attached to it.

Example 2: In MYCIN a rule based expert system (which we will discuss later), a
rule in which statements which link evidence to hypotheses are expressed as decision
criteria, may look like :

IF patient has symptoms s1,s2,s3 and s4

AND certain background conditions t1,t2 and t3 hold
THEN the patient has disease d6 with certainty 0.75

For detailed discussion on certainty factors, the reader may refer to probability theory,
fuzzy sets, possibility theory, Dempster-Shafter Theory etc.

Exercise 4

In the “Animal Identifier System” discussed above use forward chaining to try to
identify the animal called “raja”.

23
1.5 EXAMPLES OF EXPERT SYSTEMS: MYCIN, Expert Systems

COMPASS

The first expert system we choose as and example is MYCIN, which is of the earliest
developed expert systems. As another example of and expert system we briefly
discuss COMPASS.

MYCIN (An expert system)

Like every one else, we are also tempted to discuss MYCIN, one of the earliest
designed expert systems in Stanford University in 1970s.

MYCIN‟s job was to diagnose and recommend treatment for certain blood infections.
To do the proper diagnosis, it is required to grow cultures of the infecting organism
which is a very time consuming process and sometime patient is in a critical state. So,
doctors have to come up with quick guesses about likely problems from the available
data, and use these guesses to provide a treatment where drugs are given which should
deal with any type of problem.

So MYCIN was developed in order to explore how human experts make these rough
(but important) guesses based on partial information. Sometimes the problem takes
another shape, that an expert doctor may not available every-time every-where, in that
situation also and expert system like MYCIN would be handy.

MYCIN represented its knowledge as a set of IF-THEN rules with certainty factors.
One of the MYCIN rule could be like :

IF infection is primary-bacteremia AND the site of the culture is one of the sterile
sites
AND the suspected portal of entry is the gastrointestinal tract
THEN there is suggestive evidence (0.8) that bacteroid infection occurred.

The 0.8 is the certainty that the conclusion will be true given the evidence. If the
evidence is uncertain the certainties of the pieces of evidence will be combined with
the certainty of the rule to give the certainty of the conclusion.

MYCIN has been written in Lisp, and its rules are formally represented as lisp
expressions. The action part of the rule could just be a conclusion about the problem
being solved, or it could be another lisp expression.

MYCIN is mainly a goal-directed system, using the backward chaining reasoning

approach. However, to increase it reasoning power and efficiency MYCIN also uses
various heuristics to control the search for a solution.

One of the strategy used by MYCIN is to first ask the user a number of predefined
questions that are most common and which allow the system to rule out totally
unlikely diagnoses. Once these questions have been asked, the system can then focus
on particular and more specific possible blood disorders. It then uses backward
chaining approach to try and prove each one. This strategy avoids a lot of
unnecessary search, and is similar to the way a doctor tries to diagnose a patient.

The other strategies are related to the sequence in which rules are invoked. One of the
strategy is simple i.e., given a possible rule to use, MYCIN first checks all the
antecedents of the rule to see if any are known to be false. If yes, then there is no point
24
Applications of Artificial using the rule. The other strategies are mainly related to the certainty factors.
Intelligence MYCIN first find the rules that have greater degree of certainty of conclusions, and
abandons its search once the certainties involved get below a minimum threshold, say,
0.2.

There are three main stages to the interaction with MYCIN. In the first stage, initial
data about the case is gathered so the system can come up with a broad diagnosis. In
the second more directed questions are asked to test specific hypotheses. At the end of
this section it proposes a diagnosis. In the third stage it asks questions to determine an
appropriate treatment, on the basis of the diagnosis and facts related to the patient.
After that it recommends some treatment. At any stage the user can ask why a
question was asked or how a conclusion was reached, and if a particular treatment is
recommended the user can ask if alternative treatments are possible.

MYCIN has been popular in expert system‟s research, but it also had a number of
problems or shortcomings because of which a number of its derivatives like
NEOMYCIN developed.

COMPASS (An expert system)

Before we discuss this let us understand the functionality of telephone company‟s

switch. A telephone company‟s switch is a complex device whose circuitry may
encompass a large part of a building. The goals of switch maintenance are to minimize
the number of calls to be rerouted due to bad connections and to ensure that faults are
repaired quickly. Bad connections caused due to failure of connection between two
telephone lines.

COMPASS is an expert system which checks error messages derived from the
switch‟s self test routines, look for open circuits, reduces the time of operation of
components etc. To find the cause of a switch problem, it looks at a series of such
messages and then uses it expertise. The system can suggest the running of additional
tests, or the replacement of a particular component, for example, a relay or printed
circuit board.

As expertise in this area was scarce, so it was a fit case for taking help an expert
system like COMPASS (We will discuss later, how knowledge acquisition is done in
COMPASS).

1.6 EXPERT SYSTEM BUILDING TOOLS

Expert system tools are designed to provide an environment for development of expert
systems mainly through the approach of prototyping.

The expert systems development process is normally a mixture of prototyping and

other incremental development models of software engineering rather than the
conventional “waterfall model”. Although using incremental development has a
problem of integrating new functionality with the earlier version of expert system but
the expert system development environments try to solve this problem by using
modular representations of knowledge (some of the representation schemes like
frames, rule based systems etc. have already been discussed above).

Software tools for development of expert systems mainly fall into the following
Categories:

25
 Expert System Shells: These are basically a set of program and to be more Expert Systems
specific - abstractions over one or more applications programs. One of the major
examples is EMYCIN which is the rule interpreter of the famous expert system
called MYCIN (a medical diagnostic system). EMYCIN also constitutes related
data structures like knowledge tables and indexing mechanism over these tables.
Some recent versions of EMYCIN like M.4 is a very sophisticated shell which
combine the backward chaining of EMYCIN with frame – like data structures.

 High Level Programming Languages : These languages are basically used to

provide a higher level of abstraction by hiding the implementation details because
of which the programmer need not worry about efficiency of storing, accessing
and manipulating data. One of the good examples is OPS5 rule language. But
many of such languages are research tools and may not available commercially.

 Multiple programming environments : These provide a set of software modules

to allow the user to use and mix a number of different styles of artificial
intelligence programming. One of the examples is called LOOPS which combines
rule based and object – oriented approaches of knowledge representation.

We will now keep our focus on “Expert System Shells”.

1.6.1 Expert System Shells

An expert system tool, or shell, is a software development environment containing the

basic components of expert systems. Associated with a shell is a prescribed method
for building applications by configuring and instantiating these components.

Expert system shells are basically used for the purpose of allowing non-programmers
to take advantage of the already developed templates or shells and which have
evolved because of the efforts of some pioneers in programming who have solved
similar problems before. The core components of an expert systems are the knowledge
base and the reasoning engine.

A generic expert system shell is shown below :

Expert System Shell

Knowledge Knowledge Inference Explanation
Expert Acquisition Base Engine Subsystem User User
Subsystem (Reason- Inter-
ing) face

Figure 1.2: Components of Expert System Tool (Shell)

As we can see in the figure above, the shell includes the inference engine, a
knowledge acquisition subsystem, an explanation subsystem and a user interface.
When faced with a new problem in any given domain, we can find a shell which can
provide the right support for that problem, so all we need is the knowledge of an
expert. There are many commercial expert system shells available now, each one
adequate for a different range of problems. Taking help of expert system shells to
develop expert systems greatly reduces the cost and the time of development as
compared to developing the expert system from the scratch.

Let us now discuss the components of a generic expert system shell. We will discuss
about:
26
Applications of Artificial  Knowledge Base
Intelligence
 Knowledge Acquisition Subsystem
 Inference Engine
 Explanation Subsystem
 User Interface

1.6.1.1 Knowledge Base

It contains facts and heuristic knowledge. Developers try to use a uniform

representation of knowledge as for as possible. There are many knowledge
representation schemes for expressing knowledge about the application domain and
some advance expert system shells use both frames (objects) and IF-THEN rules. In
PROLOG the knowledge is represented as logical statements.

1.6.1.2 Knowledge Acquisition Subsystem

The process of capturing and transformation of potentially useful information for a

given problem from any knowledge source (which may be a human expert) to a
program in the format required by that program is the job of a knowledge acquisition
subsystem. So we can say that these subsystem to help experts build knowledge bases.

As an expert may not be a computer literate, so capturing information includes

interviewing, preparing questionnaires etc. which is a very slow and time consuming
process. So collecting knowledge needed to solve problems and build the
knowledge base has always been the biggest bottleneck in developing expert
systems.

Some of the reasons behind the difficulty in collecting information are given below :

o The facts and rules or principles of different domains cannot easily be

converted into a mathematical or a deterministic model, the properties of which
are known. For example a teacher knows how to motivate the students but
putting down on the paper, the exact reasons, causes and other factor affecting
students is not easy as they vary with individual students.

o Different domains have their own terminology and it is very difficult for experts
to communicate exactly their knowledge in a normal language.

o Capturing the facts and principles alone is not sufficient to solve problems. For
example, experts in particular domains which information is important for
specific judgments, which information sources are reliable and how problems
can be simplified, which is based on personal experience. Capturing such
knowledge is very difficult.

o Commonsense knowledge found in humans continues to be very difficult to

capture, although systems are being made to capture it.

The idea of automated knowledge capturing has also been gaining momentum. Infact
“machine learning” is one of the important research area for sometime now. The
goal is that, a computing system or machine could be enabled to learn in order to solve
problems like the way human do it.

Example (Knowledge acquisition in COMPASS).

As we already know that COMPASS is an expert system which was build for proper
maintenance and troubleshooting of telephone company’s switches.
27
Now, for knowledge acquisition, knowledge from a human expert is elicited. An Expert Systems
expert explains the problem solving technique and a knowledge engineers then
converts it into and if-then-rule. The human expert then checks if the rule has the
correct logic and if a change is needed then the knowledge engineer reformulates the
rule.

Sometimes, it is easier to troubleshoot the rules with pencil and paper (i.e., hand
simulation), at least the first time than directly implementing the rule and changing
them again and again.

Let us summarize the knowledge acquisition cycle of COMPASS :

 Extract knowledge from a human expert.

 Document the extracted knowledge.
 Test the new knowledge using following steps:

 Let the expert analyze the new data.

 Analyze the same data using pencil and paper using the documented
knowledge.
 Compare the results of the expert’s opinion with the conclusion of the
analysis using pencil and paper.
 If there is a difference in the results, then find the rules which are the cause
of discrepancy and then go back to rule 1 to gather more knowledge form
the expert to solve the problem.
Otherwise, exit the loop.

1.6.1.3 Inference Engine

An inference engine is used to perform reasoning with both the expert knowledge
which is extracted from an expert and most commonly a human expert) and data
which is specific to the problem being solved. Expert knowledge is mostly in the form
of a set of IF-THEN rules. The case specific data includes the data provided by the
user and also partial conclusions (along with their certainty factors) based on this data.
In a normal forward chaining rule-based system, the case specific data is the elements
in the working memory.

Developing expert systems involve knowing how knowledge is accessed and used
during the search for a solution. Knowledge about what is known and, when and
how to use it is commonly called meta-knowledge. In solving problems, a certain
level of planning, scheduling and controlling is required regarding what questions to
be asked and when, what is to be checked and so on.

Different strategies for using domain-specific knowledge have great effects on the
performance characteristics of programs, and also on the way in which a program
finds or searches a solution among possible alternatives. Most knowledge
representations schemes are used under a variety of reasoning methods and research is
going on in this area.

1.6.1.4 Explanation Subsystem (Example MYCIN)

An explanation subsystem allows the program to explain its reasoning to the user. The
explanation can range from how the final or intermediate solutions were arrived at to
justifying the need for additional data.

Explanation subsystems are important from the following points of view :

28
Applications of Artificial (i) Proper use of knowledge: There must be some for the satisfaction of
Intelligence knowledge engineers that the knowledge is applied properly even at the time of
development of a prototype.
(ii) Correctness of conclusions: User’s need to satisfy themselves that the
conclusions produced by the system are correct.
(iii) Execution trace: In order to judge that the knowledge elicitation is proceeding
smoothly and successfully, a complete trace of program execution is required.
(iv) Knowledge of program behavior: For proper maintenance and debugging, the
knowledge of program behavior is necessary for the programmers.
(v) Suitability of reasoning approach: Explanation subsystems are necessary to
ensure that reasoning technique applied is suitable to the particular domain.

Explanation in expert systems deals with the issue of control because the reasoning
steps used by the programs will depend on how it searches for a solution.

Explanation subsystems are also related to evaluation as by checking the outputs

produced by a system and after examining the trace of its actions performed while
reasoning, it can be decided that whether or not a system is producing the right answer
for right reasons. Without a good explanation subsystem, an expert will be unable to
judge the system’s performance or will be unable to find ways to improve it.

To give an idea of how an explanation subsystem works, let us have a simple

example:

Example:

Suppose there is a simple rule based system to diagnose the problems in a car.

Let the system has following rules :

R1: IF engine-gets-petrol AND engine-turns-over
THEN spark-plug-fault

R2: IF NOT engine-turns-over AND NOT lights-come-on

THEN battery-problem

R3: IF NOT engine-turns-over AND lights-come-on

THEN starter-not-working

R4: IF fuel-tank-has-petrol
THEN engine-gets-petrol

We also assume that the system has no initial facts.

Explanation subsystems allow the user to ask why it asked a particular question, and
how it reached some conclusion. These questions are answered by referring to the
system goals, the rules being used, and any existing problem data.

Now let us focus on the example given above to see the explanation facilities
provided, which involve a dialogue involving why and how questions and their
explanations.

System : Does the fuel tank has gas?

User : Yes.

29
 Expert Systems
System : Does the engine turns over?
User : No

System : Do lights come on?

User : Why?

System : I am trying to prove that there is a battery problem. It has been

established
that the engine does not turns over. Therefore if it is not the case that the
lights
come on, then there is a battery problem.
Is it true that the lights come on?
User : No.

System : I conclude that there is a battery problem.

User : How?

System : This follows from Rule 2:

IF NOT engine-turns-over AND NOT lights-come-on was given by the
user.

Providing such an explanation facility involves stating what rules are used in reaching
conclusions, and using these records to compose explanations like the ones above.
Giving simple explanations like those above is not very difficult, and is quite useful.

Explanation facilities in expert systems are sometimes not used, and sometimes they
are not easily accepted by their users. There are a whole lot of reasons for this. One
reason is that the explanations just reference the “surface” knowledge encoded in the
rules, rather than providing the “deep” knowledge about the domain which originally
motivated the rules but which is usually not represented. So, the system will say that it
concluded something because of rule 5, but not explain what rule 5 intends to say. In
the example given below, maybe the user needs to understand that both the lights and
the starter use the battery, which is the underlying purpose of the second rule in this
example. Another reason for the frequent failure of explanation facilities is the fact
that, if the user fails to understand or accept the explanation, the system can‟t re-
explain in another way (as people can). Explanation generation is a fairly large area of
research, concerned with effective communication i.e., how to present things so that
people are really satisfied with the explanation, and what implications does this have
for how we represent the underlying knowledge.

Explanation Subsystem in MYCIN (An overview)

MYCIN is one of the first popular expert systems made for the purpose of medical
diagnosis. Let us have a look at how the explanation subsystem in MYCIN works :
To explain the reasoning for deciding on a particular medical parameter’s or
symptom’s value, it retrieves a set of rules and their conclusions. It allows the user to
ask questions or queries during a consultation.

To answer the questions the system relies on its ability to display a rule invoked at any
point during the consultation and also recording the order of rule invocations and
associating them with particular events (like particular questions).

As the system using backward chaining so most of the questions belong to “Why” or
“How” category. To answer “Why” questions, the system looks up the hierarchy (i.e.,
tree) of rules to see which goals the system is trying to achieve and to answer “Why”

30
Applications of Artificial questions, the system must look down the hierarchy (i.e., tree) to find out that which
Intelligence sub-goals were satisfied to achieve the goal.

We can see that explanation process is nothing but a search problem requiring tree
traversal.

As MYCIN keeps track of the goal to sub-goal sequence of the computation, so it can
answer questions like:

“Why did you ask that if the stain of the organism is gram negative ?”

In response to this, the system would quote the rules which states “gram negative
staining” may be in conjunction with other conditions and would state that what it was
trying to achieve.

Similarly, if there is “How” question like:

“How do say that Organism-2 might be proteus ?”

In its reply, MYCIN would state the rules that were applied in reaching this
conclusions and their degree of certainty and what was the last question asked etc.

Thus we can see that because of the backward chaining approach, the system is able to
answer “Why” and “How” questions satisfactorily. But the rule application would not
be easy if the chains of reasoning are long.

1.6.1.5 User interface

It is used to communicate with the user. The user interface is generally not a part of
the expert system technology, and was not given much attention in the past. However,
it is now widely accepted that the user interface can make a critical difference in the
utility of a system regardless of the system’s performance.

Now as an example, let us discuss and expert system shell called EMYCIN.

1.6.1.6 EMYCIN (An expert system shell)

EMYCIN provides a domain-independent framework or template for constructing and

running any consultation programs. EMYCIN stands for “Empty MYCIN” or
“Essential MYCIN” because it basically constitutes a MYCIN system minus its
domain-specific medical knowledge. However, EMYCIN is something more than this,
as it offers a number of software tools for helping expert system designers in building
and debugging performance programs.

Some characteristics of EMYCIN are:

 It constitutes an abbreviated rule language, which uses ALGOL-like notation and

which is easier than LISP and is more concise than the English subset used by
MYCIN.
 It uses backward chaining which is similar to MYCIN.
 It indexes rules, in-turn organising them into groups, based on the parameters
which are being referenced.
 It has an interface for system designer which provides tools for displaying,
editing and partitioning rules, editing knowledge held in tables, and also running
rule sets on sets of problems. As part of system designer‟s interface, EMYCIN

31
 also included a knowledge editor (a program) called TEIRESIAS whose job was Expert Systems
to provide help for the development and maintenance of large knowledge bases.
 It also has a user interface which allows the user to communicate with the system
smoothly.

1.7 SOME APPLICATION AREAS OF EXPERT

SYSTEMS

The scope of applications of expert systems technology to practical problems is so

wide that it is very difficult to characterize them. The applications find their way into
most of the areas of knowledge work. Some of the main categories of applications or
an expert system are given below.

 Diagnosis and Troubleshooting: This class comprises systems that deduce faults
and suggest corrective actions for a malfunctioning device or process. Medical
diagnosis was one of the first knowledge areas to which ES technology was
applied, but use of expert systems for solving and diagnosis of engineered systems
has become common.

 Planning and Scheduling: Systems that fall into this class analyze a set of one or
more potentially complex and interacting goals in order to determine a set of
actions to achieve those goal. This class of expert systems has great commercial
potential. Examples include scheduling of flights, personnel, manufacturing
process planning etc.

 Process Monitoring and Control: Systems falling in this class analyze real-
time data from physical devices with the goal of noticing errors, predicting
trends, and controlling for both optimality and failure correction. Examples of
real-time systems that actively monitor processes are found in the steel making
and oil refining industries.

 Financial Decision Making: The financial services industry has also been using
expert system techniques. Expert systems belonging to this category act as
advisors, risk analyzers etc.

 Knowledge Publishing: This is a relatively new, but also potentially explosive

area. The primary function of the expert system is to deliver knowledge that is
relevant to the user‟s problem, in the context of the user‟s problem.

 Design and Manufacturing: These systems assist in the design of physical

devices and processes, ranging from high-level conceptual design of abstract
entities all the way to factory floor configuration of manufacturing processes.

1.8 SUMMARY

In this unit, we have discussed various issues in respect of expert systems. To begin
with, in section 1.2 we define the concept of „expert system‟. In view of the fact that
an expert system contains knowledge of a specific domain, in section 1.4, we discuss
various schemes for representing the knowledge of a domain. Section 1.5 contains
examples of some well-known expert systems. Tools for building expert systems are
discussed in section 1.6. Finally, applications of expert systems are explained in
section 1.7.
32
Applications of Artificial
Intelligence 1.9 SOLUTIONS/ANSWERS

Ex. 1) In this case, it is not the same „strike‟ action but two strike actions which are
involved in the sentence. Therefore, we use „strike‟ to denote a generic action
of striking whereas „strike-1‟ and „strike-2‟ are its instances or members.
Thus, we get the semantic network.

Strike-1 instance Strike instance Strike-2

agent
agent
past of
object
Nita Mohan struck X

Mother-of

past of

struck

Ex. 2) (date
(day (integer (1….31)))
(month (integer (1…..12)))
(year (integer (1…..10000)))
(day-of-the-week (set (Mon Tue Wed Thu Fri Sat Sun)))
(procedure (compute day-of-the-week (day month year))),

Where the procedure day-of-the-week takes three arguments.

Some problems in Semantic Networks and Frames:

There are problems in expressing certain kinds of knowledge when either semantic
networks or frames are used for knowledge representation. For example, it is difficult
although not impossible to express disjunctions and hence implications, negations,
and general non-taxonomic knowledge (i.e., non-hierarchical knowledge) in these
representations.

Ex. 3) In order to translate in PL, let us use the symbols:

ML: There is a Moral Law
SG: Someone Gave it (the word ‘it’ stands for moral law)
TG: There is God.
Using these symbols, the statements (i) to (iv) become the formulae (i) to (iv) of PL
as given below:
(i) ML
(ii) MLSG
(iii) SGTG and
33
 (iv) TG Expert Systems

Applying Modus Ponens to formulae (i) and (ii) we get the formula
(v) SG
Applying Modus Ponens to (v) and (iii), we get
(vi) TG
But formula (vi) is the same as (iv), which is required to be established. Hence the
proof.

Ex. 4) Initially, the working memory contains the following assertions :

(has-hair raja) representing the fact “raja has hair”

(big-mouth raja) representing the fact “raja has a big mouth”
(long-pointed-teeth raja) representing the fact “raja has long pointed
teeth”
(claws raja) representing the fact “raja has claws”

We start with one of the assertions about “raja” from the working memory.

Let us choose the first assertion : (raja has-hair)

Now we take this assertion and try to match the antecedent part of a rule. In the rule
2, the antecedent part is satisfied by substituting “raja” in place of X. So the
consequent (mammal X) is established by replacing “raja” in place of X.

The working memory is now updated by adding the assertion (mammal raja).
So the working memory now looks like:

(mammal raja)
(has-hair raja)
(big-mouth raja)
(long-pointed-teeth raja)
(claws raja)

Now we try to match assertion (mammal raja) to the antecedent part of a rule. The
first rule whose antecedent part supports the assertion is rule 3. So the control moves
to the second part of rule 3, which says that (eats-meat X), but this is not found in any
of the assertions present in working memory, so rule 3 fails.

Now, the system tries to find another rule which matches assertion (mammal raja), it
find rule 4 whose first part of antecedent supports this. So the control moves to the
second part of the antecedent in rule 4, which says that something i.e., X must have
pointed teeth. This fact is present in the working memory in the form of assertion
(long-pointed-teeth raja) so the control now moves to the third antecedent part of rule
4 i.e., something i.e., X must have claws. We can see that this is supported by the
assertion (claws raja) in the working memory. Now, as the whole antecedent of rule 4
is satisfied so the consequent of rule 4 is established and the working memory is
updated by the addition of the assertion (carnivorous raja), after substituting “raja” in
place of X.

So the working now looks like:

(carnivorous raja)
34
Applications of Artificial (mammal raja)
Intelligence
(has-hair raja)
(big-mouth raja)
(long-pointed-teeth raja)
(claws raja)

Now in the next step, the system tries to match the assertion (carnivorous raja) with
one of the rules in working memory. The first rule whose antecedent part matches this
assertion is rule 6. Now as the first part of the antecedent in rule 6 matches with the
assertion, the control moves to the second part of the antecedent i.e., X has dark spots.
There is no assertion in working memory which supports this, so the rule 6 is aborted.

The system now tries to match with the next rule which matches the assertion
(carnivorous raja). It finds rule 8 whose first part of antecedent matches with the
assertion. So the control moves to the second part of the antecedent of rule 8 which
says that something i.e., X must have big mouth. Now this is already present in the
working memory in the form of assertion (big-mouth raja) so the second part and
ultimately the whole antecedent of rule 8 is satisfied.

And, so the consequent part of rule 8 is established and the working memory is
updated by the addition of the assertion (lion raja), after substituting “raja” in place of
X.

The working memory now looks like:

(lion raja)
(carnivorous raja)
(mammal raja)
(has-hair raja)
(big-mouth raja)
(long-pointed-teeth raja)
(claws raja)

Hence, as the goal to be achieved i.e., “raja is a lion” is now part of the working
memory in the form of assertion (lion raja), so the goal is established and processing
stops.

1.10 FURTHER READINGS

1. Russell S. & Norvig P., Artificial Intelligence A Modern Approach (Second

Edition) (Pearson Education, 2003).

2. Patterson D. W: Introduction to Artificial Intelligence and Expert Systems

(Prentice Hall of India, 2001).

35