TRANSFERRING PROBLEM SOLVING STRATEGIES FROM

THE EXPERT TO THE END USERS

Supporting understanding

Anne Håkansson

Department of Information Science, Computer Science, Uppsala University, Box 513, SE-751 20, Uppsala, Sweden

Keywords: Knowledge Management S

ystems, Knowledge Sharing, Knowledge Dissemination, E-Learning, Reasoning

Strategies, Visualization, Graphical Diagrams and Modeling Formalisms

Abstract: If knowledge sharing between people in an organisation is to be encouraged, new types of systems are

needed to transfer domain knowledge and problem-solving strategies from an expert to the end users and,

thereby, make the knowledge available and applicable in a specific domain. If it is to be possible to apply

the knowledge in the organisation, the systems will need a means of illustrating the reasoning strategies

involved in interpreting the knowledge to arrive at the conclusions drawn. One solution is to incorporate

different diagrams in knowledge management systems to assist the user to comprehend the reasoning

strategies and to better understand the knowledge required and gained. This paper describes the manners by

which knowledge management systems can facilitate transfer of problem-solving strategies from a domain

expert to different kinds of end users. With this objective in mind, we suggest using visualization, graphical

diagrams and simulation in conjunction to support the transfer of problem-solving strategies from a domain

expert to the end users. Visualization can support end users, enabling them to follow the reasoning strategy

of the system more easily. The visualization discussed here includes static and dynamic presentation of the

rules and facts in the knowledge base that are used during execution of the system. The static presentation

illustrates how different rules are related statically in a sequence diagram in the Unified Modeling Language

(UML). The dynamic presentation, in contrast, visualizes rules used and facts relevant to a specific

consultation, i.e., this presentation depends on the input inserted by the users and is illustrated in a

collaboration diagram in the UML. Utilising these diagrams can support the sharing and reuse of the

knowledge and strategies used for handling routine tasks and problems more efficiently and profitably

whilst minimizing potential for loss of knowledge. This is important when experts are not available on the

spot. These diagrams can also be used for the organisation and the disseminating of knowledge by locating

experts in an organisation, which is important when these are to be relocated in large organisations or

geographically distributed.

1 INTRODUCTION

Knowledge management refers to an organisation’s

ability to learn from its environment and to

incorporate knowledge into its business processes

(Laudon & Laudon, 2002). This provides

instruments with which to optimise the control and

the management of crucial production factors and

aims at preventing bottlenecks of the kind that arise

when information is not transferred smoothly within

an organisation. The essence is the organisation of

processes through which knowledge is developed

and distributed to those who need it. It also involves

making knowledge accessible for future use by the

whole organisation and the combining of different

knowledge areas (Liebowitz & Wilcox, 1997).

Knowledge management is embodied by a set of

ocesses developed in an organisation to gather,

organise, refine and disseminate knowledge (Awad

& Ghaziri, 2004). In this, information technology

plays an important role. For example, a knowledge

management system can enable the creation, storage,

maintenance, and dissemination of knowledge, it can

optimise learning and protecting whilst allowing it to

be shared between people in an organisation

(Laudon & Laudon, 2002). It enables people to act

in an informed manner when a new source of

information becomes available and to deal with the

information in a beneficial way.

Håkansson A. (2005).

TRANSFERRING PROBLEM SOLVING STRATEGIES FROM THE EXPERT TO THE END USERS - Supporting understanding.

In Proceedings of the Seventh International Conference on Enterprise Information Systems, pages 3-10

DOI: 10.5220/0002526100030010

 SciTePress

When a user is operating a knowledge

management system, the knowledge that should be

transferred is that possessed by an expert, i.e. it is

domain knowledge and will incorporate problem

solving strategies that need to be passed on to the

end users. In knowledge management systems, the

domain knowledge is usually expressed in terms of

facts, rules, concepts, relationships, assumptions and

tasks (Tansley & Hayball, 1993). The problem

solving or so-called reasoning strategy usually

involves deductive or inductive strategies but could

also involve a combination of these (Durkin, 1994).

Knowledge management systems must allow

organisations to store and access information more

efficiently (Awad & Ghaziri, 2004). Moreover, these

systems must document how decisions are reached.

This knowledge needs to be distributed within the

organisation. Therefore, domain knowledge and

reasoning strategies are to be transferred to end users

to support the sharing of knowledge between people.

New types of knowledge management systems

are needed to display the contents of a system; these

can work as knowledge systems with educational

facilities. In that connection, the system has to

provide several different views of the knowledge to

support the different end users. Moreover, the

system has to have a procedure for illustrating the

strategies involved in the interpretation of the

knowledge.

One way of transferring domain knowledge and

reasoning strategies from a domain expert to an end

user via a knowledge management system is by

visualizing the knowledge and the strategies. To this

end, we use conceptualization, i.e., we exploit

concepts and relationships together with

visualization. Concepts correspond to facts, and

relationships are equivalent to rules and heuristics.

Since it concerns strategies, the presentation needs

to be illustrated with by a model showing stepwise

execution. A candidate for this modeling language is

the visual modeling language Unified Modeling

Language (UML) (Booch et al., 1999), particularly

since UML has become a standard for working with

software-intensive systems (Jacobson et al., 1998).

UML provides several types of graphic diagrams

that can be utilised for inserting, modifying and

tutoring the domain knowledge, as well as to

demonstrate reasoning strategies (Håkansson, 2001).

These diagrams can also be used to generate

knowledge about static and dynamic domain

knowledge and to inform the expert and the end

users about the system’s processing (Håkansson,

2003:b).

For static presentation we use the sequence

diagrams of the UML. These illustrate the

interpretation of the knowledge base by displaying

the time sequences conducting the relations between

the rules, or heuristics, i.e., it can demonstrate how

different parts interact with each other. It also

illustrates how different rules are related to other

rules, and how these rules are dependent on each

other.

For dynamic presentation we use the

collaboration diagrams of the UML since these

display how different parts collaborate with each

other. Dynamic presentation depends on the input

the users insert into the diagram, i.e., the diagram is

dynamic in the sense that it changes with the input,

and visualizes the rules and relationships according

to the inserted facts. Thus, it visualizes the system’s

reasoning strategy, which changes with the input.

Dynamic information is relevant since the rules that

are used during an interpretation depend on the

information that is supplied by the end users

(Håkansson 2003:a; Håkansson 2003:b), and it is

usually the case that the end users supply the

additional information in these kinds of systems.

These different diagrams can be used for several

different purposes. For instance, they can be used to

solve tasks or problems when suitable opportunities

of getting support from experts will not arise. This

requires that the diagrams contain all the knowledge

necessary to solve a problem or task.

These diagrams can also be used to locate

experts in large organisations or at different

geographical places who will be needed to solve a

particular task. A combination of the questions

appropriate to a particular task or problem and

comprehensive rules can pinpoint that expert to be

consulted to provide particular support or to find a

solution.

The next section is an overview of related work;

this is followed by a discussion of transferring

problem solving and of the notion of

conceptualization. The following sections contain

descriptions of static and dynamic presentations

where statically related rules are illustrated in

sequence diagrams and dynamically related rules are

illustrated in collaboration diagrams.

2 RELATED WORK

Within knowledge management, knowledge maps

can be used to explore and solve problems

(Liebowitz, 2001). Some of these maps are

organisational maps and semantic networks.

Organisational maps can link people’s interactions

by departments in the organisation, link expertise or

knowledge areas to experts, or relate available

knowledge areas to those that are needed or missing.

The semantic network links different knowledge

areas by means of the relationships between them. In

ICEIS 2005 - ARTIFICIAL INTELLIGENCE AND DECISION SUPPORT SYSTEMS

our work, we locate where expertise or knowledge

areas can be found by examining the rules that

pinpoint the expertise that is needed to handle a

problem.

Different kinds of maps can be used to support

analyse and structure knowledge graphically,

examples of these being cognitive maps, inference

networks, flowcharts and decision trees (Durkin,

1994). Cognitive maps present domain knowledge

using nodes for concepts and objects, and links for

relationships between these. Inference networks are

also used to represent domain knowledge through

the production rules in a system, with nodes and

links providing “

AND” and “OR” branches.

Flowcharts, on the other hand, can be used to

represent reasoning strategies since they present

sequences of steps that will be performed. These are

composed of blocks with the execution order being

specified and “

YES” and NO” branches being

presented. Decision trees can represent reasoning

strategies since they use graphic presentations of

problem search spaces, composed with nodes and

arc linking related nodes. The arc can have any

value, e.g., “

LOOSE”, “>12”, and “BAD”. In our

approach, graphic diagrams are used for modeling

domain knowledge and reasoning strategy but since

the maps mentioned above suffer from problems

because they cannot cope with large systems, we

will use the diagrams developed in UML.

UML is usually used for modeling object-

oriented systems, but it can also be used for

modeling other types of systems, such as rule-based,

frames and constraint-based ones (Schreiber et al.,

2001; Håkansson, 2001; Helenius, 2001; Cranefield

et al., 2001; Renker et al., 2002).

UML diagrams are used in CommonKADS to

build knowledge-based systems in an object-oriented

fashion. Diagrams can help to model the state of a

system over a period of time and to model the

dynamic behaviour of the system, providing an

image of the sequence of events and assisting with

the decision-making (Schreiber et al., 2001).

Diagrams are also used to clarify the context, from

which the information has come, for the task

analysis and the structure of objects handled in a

task. Moreover, diagrams can be incorporated to

present the actors and the services (or use-cases) and

to include additional chunks of information that are

difficult to model, e.g., large or complex systems

(Schreiber et al., 2001).

In our approach, however, we will apply UML

diagrams to rule-based knowledge management

systems that have been developed in a declarative

fashion. This affects the UML’s diagrams, since they

cannot be used in their original form as they are to

be used in CommonKADS. In its current form,

UML is not directly applicable for modeling

knowledge in systems that are rule-based, however

UML can easily be adapted to knowledge

management systems by utilising rules in the

knowledge base.

3 TRANSFERRING PROBLEM

SOLVING

If the end users are to be provided with adequate

support, an understanding of how the problem

solving strategies work has to be passed on to the

users. This requires that the domain knowledge,

which is used to construct the system’s reasoning,

and the strategies are explicitly described in the

system.

The domain knowledge can be expressed in the

form of declarative and semantic knowledge, which

in turn can be expressed by using conceptualization.

Conceptualization is the use of concepts and

relationships applied to the domain knowledge

(Håkansson, 2003:c). These concepts and

relationships are then presented as facts and rules or

heuristics in the system to provide the declarative

knowledge. The concepts can also express semantic

knowledge, provided that these concepts are

described with words and used in a well-defined

context.

The expert’s problem-solving strategy is

presented as a reasoning strategy, often in the form

of deductive reasoning and/or inductive reasoning,

both of which are common in these systems. The

reasoning strategy is the interpretation of the

system’s knowledge, and in the process of

interpretation facts, rules and heuristics are

examined to reach the conclusions.

During the interpretation, the system will gather

the specific concepts, including facts, and

relationships between rules and heuristics that led to

the different conclusions. These concepts and

relationships constitute paths with knowledge. These

paths can be considered to be simulation strategies

illustrating how the problem solving is used to solve

a particular problem. Thus, simulation, together with

explicit reasoning strategies and conceptualization

can support different end users to help then to

understand the problem solving strategy and the

domain knowledge (Mayiwar, & Håkansson, 2004).

4 CONCEPTUALIZATION

Some kind of knowledge representation has to be

used to represent the knowledge and strategies in a

system. In this work, we use facts and production

TRANSFERRING PROBLEM SOLVING STRATEGIES FROM THE EXPERT TO THE END USERS - Supporting

understanding

rules. However, as contents of the system grow it

becomes difficult to explore the system because

there is too much internal complexity. The

complexity of the production rules can be attributed

to the information in the conclusion-part and to the

internal rules specified in the premises-part. The

premises-part is usually comprised of relationships

to other rules and facts, and thus constitutes a

complex knowledge space. This complexity

influences the way the search of the bases is

performed and affects the information retrieval. If a

small number of rules with little internal content are

used, the processing of the rules is effortless, but

otherwise the processing will be time-consuming

and labour intensive. This needs to be improved and

simplified if one is to be able to handle large bases.

A comprehension of what the rules achieve can

be obtained by examining what happens when it is

executed. This action can be labelled by adding

some semantic information, i.e., a elucidate concept.

Thereby, these concepts can be applied to rules with

the intention of grasping the meaning of the rules.

The users can define their own concepts and then

utilise and apply these self-defined concepts to the

rules. An assigned concept corresponds to drawing a

conclusion about the role of a rule by applying a

semantic meaning to that rule, which corresponds to

gaining an understanding of what the application of

that rule achieves.

The user is the person who develops the

knowledge management system, decides the relevant

similarity between the rules and clusters the rules

(Murphy & Pazzani, 1994; Wiratunga & Craw,

2000) that together accomplish a certain task. The

application of concepts can decrease the search for

rules dealing with similar tasks or topics and, in so

doing, decrease redundancy. Moreover,

conceptualization by using clustering can support

the definition of concepts on a higher level of

abstraction and recognising similar rules at this level

may allow them to be generalised.

5 STATIC AND DYNAMIC

PRESENTATION

When a knowledge management system is consulted

or when its operation needs to be understood, the

order in which the different parts of the sequence

interact with each other needs to be understood. This

specification of the order in which the relevant rules

are applied is called the call sequence. It is important

because the sequence in which the rules are applied

determines time sequence of the execution order,

i.e., when an operation is performed. The structure

of the content will already have been determined

since this is imposed during the development of the

system. One of the challenges of the presentation is

to illustrate how different rules are related. Usually

one single rule is related to several others and,

therefore, becomes dependent on them. Thus, these

connections or relations are also important since the

output depends on them too (Håkansson 2003:a;

Håkansson 2003:b).

The contents of the knowledge base, comprised

of rules and relationships, facts and conclusions are

illustrated to obtain a static picture of the

knowledge. This illustration should be a pure

presentation of the contents of the knowledge base at

a given time without any external influence from the

users. The sequence presentation is used to obtain an

overview of the contents by illustrating the rules and

their relationships in a sequence diagram. The

diagram can be used to check the static relationship

between rules. As mentioned, static information

reveals the manner in which rules are connected to

the other rules in the knowledge base.

A dynamic presentation of the knowledge

incorporates the user-supplied facts in a

collaboration diagram with relations showing the

flows over time as computations are performed. In

this diagram, the dynamic presentation of the rules

depends on the input of the users. It is dynamic in

the sense that it changes with the inputs, and it

visualizes the rules and their relationships in

accordance with the inputs. Since the collaboration

diagrams show how different rules and facts are

invoked, they provide a sequential illustration of the

steps that are involved in the interpretation made to

arrive at a specific conclusion.

The dynamic presentation can show the entire

execution through simulation of the reasoning

strategy of the system. The presentation is a step-by-

step performance of the system’s execution of its

rules. As a starting point, the diagram takes the

inputs as facts into the diagram and then processes

the rules, stepwise, until a conclusion is reached. In

this way, the end users can follow the reasoning

strategy reproduce a particular session. For

educational purposes the end users can carry out

experiments by changing the inputs and then check

the new result. It is possible to simulate strategies as

long as the inputs and rules lead to a conclusion.

Thus end users can comprehend a strategy adopted

and participate in simulations (Håkansson, 2003:c).

Now that we have introduced concepts and

terms, it is possible to examine an example to see

how these ideas work in practice. The example

selected is childhood diseases, which is collected

from a lexicon about diseases and includes measles,

rubella and chicken pox or allergic purpura and

cerebral membrane inflammation.

ICEIS 2005 - ARTIFICIAL INTELLIGENCE AND DECISION SUPPORT SYSTEMS

6 STATICALLY RELATED RULES

IN SEQUENCE DIAGRAMS

As mentioned above, sequence diagrams provide a

static image of the rules and the relationships

between rules, thereby revealing how the rules are

connected in the knowledge base. Call sequences

reveal the order in which the different parts of the

sequence interact with each other, thereby

facilitating interpretation.

The order in which the rules and facts are

incorporated within each rule is irrelevant. This

means that the rules do not necessarily have to be

described first as they have been in this example.

Facts are obtained from the end user’s input and

each fact is the answer to a specific question being

answered. Of course, when it is known relevant facts

may already have been implemented in the system.

A sequence diagram presents the interpretation

by displaying time sequences between the rules, and,

as can be seen in figure 1, it demonstrates how

different parts interact with each other. It illustrates

how different rules are related to other rules, and

how these rules are dependent on each other. Thus,

the sequence diagram can be utilised to determine

the behaviour of the system by investigating its

performance.

However, displaying all rules in a system, in

which, typically, there will be usually several

hundreds, would be uncontrollable. Therefore,

instead of only visualizing rules at the lowest and

most concrete level, the diagrams can allow the

visualization of the rules’ structures at different

levels of abstraction.

The rules in a knowledge base for childhood

diseases are used to illustrate the employment of

concepts and their relations. The following rules are

to be implemented in the system, which is the formal

modeling of the domain knowledge:

rule(3, symptoms _object, symptoms_text):-

reply(size_rash, =, ‘Yes'),

reply(several_ symptoms, =,'No' ),

reply(swelling_back, =,'No' ).

rule(7,non_conclusion_object,contact_doctor_

text):-

(check(symptoms _object, symptoms_text);

reply(size_rash, =, ‘No)),

reply(rash_blister, =,'No'),

reply(one_red_spot, =,'No').

This example illustrates two different rules. The

rule “symptoms_object” (rule 3) contains three

different facts (replies): the “size_rash” with the

value “Yes” (the meaning of this value is that the

answer is yes to the question concerning the size of

the rash), the “several_symptoms” with the value

“No” and the “swelling_back” that has the value

“No”. In the other rule, “non_conclusion_object”

(rule 7) uses another rule (check in line 2) by

referring to the rule “symptoms_object” (i.e., rule 3).

It also has the facts “size_rash” where the value is

“No” meaning that there is no rash of the size

specified, “rash_blister” with the value “No” and

“one red spot” which has also the value “No”. As

can be seen in this case, the rule includes an “or-

clause” (i.e., a “;” at the end of line 2), which means

that either the rule “symptoms_object” or the fact

“size_rash” can be used during the consultation.

check rule

r: Non conclusion object

reply fact

check rule

q: Several symptoms

Present

conclusi on

q:Size rash

q:Swelling ba ck

reply fact

Yes

q: Size rash

q: R ash bl is ter

re ply fac t

q: One r ed spot

r: Symptoms object

Rule 3

reply fact

c: Contact doctor object

Figure 1: A sequence diagram including concepts applied

to a knowledge base.

A problem arises when using UML’s sequence

diagrams because there is a clumsy facility for

presenting the rules containing or-clauses, and or-

clauses are often found in rules. Because of this, the

diagram needs to be modified to support these

clauses and, in Figure 1, the or-clauses are marked

with arrows together with the word “or” labelling

the branches. In addition, the diagrams do not

support not-clauses. The not-clause is illustrated as a

cross in the figure to mark the box that cannot be

satisfied, i.e., “Not reply fact” or “Not check rule”

meaning that the response to the question or rule is

negative.

Moreover, the “Contact doctor text” is a

conclusion that is to be presented to the end users on

occasions where this conclusion is reached, the

system fetches the corresponding text from a

conclusion base. The text presented in this case is:

“If you cannot make any diagnosis by using this

schema, you should contact a doctor”.

In a sequence diagram, each object (illustrated as

a square) is an instance of a class. To mark this the

name of the object is underscored. Sometimes the

initial of the name of the class is also used. In this

diagram, the rules, facts and conclusion can also be

seen as objects since they are either the class

TRANSFERRING PROBLEM SOLVING STRATEGIES FROM THE EXPERT TO THE END USERS - Supporting

understanding

question (facts), the class rule or the class

conclusion. The underscore is omitted, but the

initials (q, r and c in figure 1) are used to make the

information in the diagram more easily digestible.

By using the diagrams it is possible to identify

the expertise that is required for a particular

problem. The conclusion of the session points out

where in the organisation one should search for the

expertise.

6.1 Packages

As the complexity of the diagrams is to great, steps

need to be taken to reduce it. The approach adopted

uses the notion of packages can be utilised in the

system as a means for encapsulating several rules

and facts included in a rule. The package facility,

which is also a UML notation for organising

elements into groups, facilitates folding knowledge

that is not currently relevant and unfolding packages

containing knowledge that is. These packages are

only used for rules since it is unlikely that packages

for facts will have any substantial impact on the

screen space. Moreover, the packages can be nested

within other packages, which means that a system

may be represented by a single high-level package.

In folding the rule, the user who is developing

the system must be confident of the contents of that

rule, unless automatic verification or validation tools

are implemented in the system. That is, the user

must be aware if automatic verification or validation

tools have been implemented in the system because

folding a rule in which they have been implemented

could be at the expense of introducing verification

and validating problems.

6.2 Change the Execution Order of

the Rules

The adoption of different strategies generally has a

marked effect upon the performance characteristics

of programs. The strategies determine the manner in

which a program searches for a solution. By

visualizing the interpretation, the reasoning strategy

of the system becomes more perceptible. With the

help of a diagram showing the rules, it should be

possible to change the execution order of the rules

and, thereby, examine the reasoning strategy but also

experiment with the execution order.

A domain expert might prefer not to reason by

starting with a conclusion and working backwards to

find a solution, as one does in backward chaining.

Instead, the expert may start with several facts in an

attempt to find a solution through forward chaining,

or, more probably, it might be decided to use a

mixed reasoning strategy. The reasoning may start

out with some facts and then use a hypothesis or

theory to find a solution. For instance, it is possible

to form a theory about a project and start to check

whether the premises that would be required to

satisfy the theory (or hypothesis) are valid. The

result would be displayed as a sequence diagram, a

diagram that allows both facts and rules to be

inserted, yielding different strategies.

With aid of such a diagrams, end users can be

able to learn how to solve a problem. The diagrams

can make it possible to experiment with the strategy

by placing the facts and rules at the lifeline

corresponding to the strategy and then moving them

back and forth. Changing the strategy will affect the

reasoning, which will give different results, because

of the different application of the facts and rules.

7 DYNAMICALLY RELATED

RULES IN COLLABORATION

DIAGRAMS

Instead of using one sequence diagram for several

purposes, dynamic knowledge can be presented in

what is known as a collaboration diagram by

incorporating user-supplied facts in the diagram.

This collaboration diagram can be used to modify

the execution order or the reasoning strategy of the

system by illustrating how these rules are related

dynamically.

Dynamic information is important since the rules

that are used during an execution depend on the

information that is supplied by the end users. User-

supplied facts are incorporated in a collaboration

diagram in which the relations show the flows that

are made over time to perform computations, and

that, therefore, illustrate the dynamic changes. In

such a diagram, the dynamic presentation of the

rules depends on the inputs the users insert into the

diagram. It is dynamic in the sense that it changes

with the inputs, and it visualizes the rules and their

relationships according to the inputs. Since

collaboration diagrams show how different rules and

facts are invoked, they give a sequential

demonstration of the steps that are involved in

arriving at a specific conclusion.

It is difficult to control and gain an overview of

relations between different parts: between the input,

the output and the rules. As demonstrated in Figure

2, the interrelation between these parts shows how

they are linked. For example, to reach the conclusion

“Contact doctor text”, the inputs, “No rash blister”,

“No rash size” and “No one red spot”, have been

inserted into the diagram. Then the rules, “Non

ICEIS 2005 - ARTIFICIAL INTELLIGENCE AND DECISION SUPPORT SYSTEMS

conclusion object” (Rule 7) and “Symptoms object”

(Rule 3) are used with the facts “Size rash”= “No”,

“Rash blister”= “No” and “One red spot” = ”No”.

call

1:check rule

r: Non conclusion object

2:check answer

r: Symptoms object

“Size rash” is “No”

5:present conclusion

c: Contact doctor text

no rash blister

no rash size

not one red spot

“Rash blister” is “No”

3:check answer

4:check answer

“One red spot” is “No”

Figure 2: A dynamic presentation of rules

In this diagram, the initials corresponding to the

name of the class have not been displayed for the

questions because these questions are displayed with

an alternative answer. This can, of course, be

included if the users prefer to use the initials.

Since the collaboration is dynamic, it is possible

to check the result against the input being changed.

By using collaboration diagrams, it may be easier to

get an overview of the entities in the sets.

A collaboration diagram makes it quite easy to

see where the rules or facts do not satisfy the inputs.

By changing the inputs until a particular fact is

satisfied, the end users can experiment with inputs to

the system and use the diagram to assist in learning.

Of course, the end users can change other inputs as

well. It should be noted that another way to present

this diagram is to display the complete diagram, as

in the first case, but to mark the non-satisfied box.

An analogy to company reports can demonstrate

difference between static and dynamic. In annual

reports there are two parts: the cashflow statement

and balance sheet. The cashflow statement shows

the movement of cash throughout the year and,

therefore, enables one to see what changes have

taken place, how the company runs, how the core

business is operating, and so on. The balance sheet

gives a frozen image of the state of the company at

one point in time and which, because companies are

obliged to produce these accounts annually, enables

one to compare the situation with previous years’

figures.

8 CONCLUDING REMARKS

Transferring problem solving strategies from an

expert to end users via a knowledge management

system is accomplished by conceptualization and by

visualization in the form of graphical diagrams.

Conceptualization is applied to on top of rules to

cope with the domain knowledge and the reasoning

strategy. The internal contents of the system are

dealt with by applying concepts in diagram-form, in

addition to which, including relationships facilitates

the handling of reasoning strategies. To transfer the

knowledge, it is necessary to include visualization

for presenting and understanding the problem

solving strategy to ensure that the end users

comprehend where the knowledge comes from.

The diagrams use concepts that correspond to the

rules. Thus, instead of presenting a rule’s physical

structure, a concept with semantic meaning is

applied to a rule. In these systems, the notion of a

concept is expected to grasp the semantics of a rule

and to convey a meaning of the operation it brings

about. Such semantics can be utilised to change how

the end users’ comprehension of the knowledge,

enabling them to understand the result of following

the different paths from a semantic meaning point of

view. Thus, diagrams can be used to explain how the

order the interpreter traverses the knowledge base to

reach a particular conclusion.

Utilising concepts and visualizing these in a

sequence and collaboration diagrams can illustrate

the rules and their relationships, in a static and

dynamic manner. Static presentation refers to the

visualization of the actual contents of the system,

here, a system promote understanding of the

reasoning strategy. Dynamic presentation, in

contrast, depends on the input the user makes to the

diagram, and thus, it is dynamic in the sense that it is

changes with the input, and is visualizes the

concepts and relationships corresponding to a

particular conclusion. Thus, it is visualizing the

system’s reasoning strategy that is visualized and

this changes depending on the particular situation or

task to be solved.

Domain knowledge and problem solving

strategies are of great importance for improving

domain knowledge and clarifying the strategies.

Each advance in the understanding of problem

solving and learning processes provides new insights

about the ways in which a learner can usefully be

supported. The systems under investigation here

simulate human reasoning and judging capabilities

by accepting knowledge from an external source and

accessing stored knowledge, applying a reasoning

process to solve problems.

If a transfer of knowledge is to be realized,

knowledge not only needs to be sent to a recipient,

but also to be absorbed and put to use (Andersson,

2000). Thus, if the knowledge and the strategies,

extracted from a system, can satisfy the users’

different learning styles then the knowledge can be

absorbed. By visualizing static reasoning strategies,

such as deductive reasoning of declarative

knowledge through the use of concepts, we believe

that people can learn to understand the problem

solving strategies. Declarative knowledge, such as

TRANSFERRING PROBLEM SOLVING STRATEGIES FROM THE EXPERT TO THE END USERS - Supporting

understanding

statements, is provided in the diagrams as well as

semantic knowledge, such as meanings, since the

diagram uses concepts to capture semantic notions at

different levels of abstraction. The declarative

knowledge and semantic knowledge can be used by

end users, who learn in different ways, for example,

verbal-linguistic intelligence, semantic knowledge

for logical-mathematical intelligence and

visualization with concepts for visual-spatial

intelligence (Mayiwar, & Håkansson, 2004).

Simulation of the dynamic behavior of an

interactive execution (or session) with the system is

another means of providing support to end users.

Visualizing procedural knowledge, i.e., step-by-step

execution, together with building student models can

support the types of intelligence mentioned above.

More work is needed to analyse the extent to

which the sequence and collaboration diagrams can

be supportive during learning and when changing of

the reasoning strategy. This may require illustration

of the relationships between certain rules by

simulating the execution order that is used to reach a

specific conclusion. Simulation will show how the

rules and facts contribute to the reasoning and,

thereby, support the development as well as the

consultation with the system.

Finally, more work is needed to analyse the

degree to which the end users can benefit from these

diagrams since they can learn to use a the strategy

by examining the reasoning followed. Moreover, it

is important to check whether they are able to

experiment with the facts and rules used by the

reasoning strategy to reach alternative conclusions.

REFERENCES

Andersson, K., 2000. Knowledge Technology Applications

for Knowledge Management, PhD thesis, Department

of Information Science, Computer Science Division,

Uppsala University, Sweden.

Awad, E.M., & Ghaziri H.M., 2004. Knowledge

Management. Pearson Education, Inc., Upper Saddle

River, New Jersey.

Booch, G., Rumbaugh, J., & Jacobson, I., 1999. The

Unified Modeling Language User Guide. Addison

Wesley Longman, Inc.

Durkin, J., 1994. Expert System Design and Development.

Prentice Hall International Editions. Macmillian

Publishing Company, New Jersey.

Helenius, E., 2001. UML and Knowledge Engineering of

Frame Based Knowledge Systems. Master Thesis,

Department of Information Science, Computer

Science Division, Uppsala University, Sweden.

Håkansson A., 2001. UML as an approach to Modelling

Knowledge in Rule-based Systems. (ES2001) The

Twenty-first SGES International Conference on

Knowledge Based Systems and Applied Artificial

Intelligence. Peterhouse College, Cambridge, UK

December 10

-12

Håkansson, A., 2003:a. Graphic Representation and

Visualisation as Modelling Support for the Knowledge

Acqusition Process. PhD thesis, Department of

Information Science, Computer Science Division,

Uppsala University, Sweden. ISBN 91-506-1727-3.

Håkansson, A., 2003:b. Supporting Illustration and

Modification of the Reasoning Strategy by

Visualisation. (SCAI’03) The Eighth Scandinavian

Conference on Artificial Intelligence, Bergen,

Norway, November 2

-4

Håkansson, A., 2003:c. Visual Conceptualisation for

Knowledge Acquisition in Knowledge Based Systems.

Accepted in: Frans Coenen (ed.): Expert Update

(SGAI) Specialist Group on Artificial Intelligence,

ISSN 1465-4091.

Jacobson, I. Rumbaugh, J. & Booch, G., 1998. The

Unified Modeling Language User Guide. Addison-

Wesley, USA.

Laudon, K., & Laudon, J., 2002. Management Information

Systems: Managing the Digital Firm. Prentice-Hall,

Inc., Upper Saddle River, New Jersey, 7

edition.

Liebowitz J., & Wilcox, L., 1997. Knowledge

Management and Its Integrative Elements. CRC Press,

LLC 2000 Corporate Blvd., N.W., Boca Raton,

Florida.

Liebowitz J., 2001. Knowledge Management: Learning

from Knowledge Engineering. CRC Press, LLC 2000

Corporate Blvd., N.W., Boca Raton, Florida.

Mayiwar, N. & Håkansson, A., 2004. Considering

Different Learning Styles when Transferring Problem

Solving Strategies from Expert to End Users.

(KES'2004) 8th International Conference on

Knowledge-Based Intelligent Information &

Engineering Systems, Wellington, New Zealand,

September 22

-24

Murphy, P. & Pazzani, M., 1994. Revision of Production

System Rule-Bases, Machine Learning: Proceedings

of the Tenth International Workshop.

Schreiber, G., Akkermans, H., Anjewierden, A., de Hoog,

R., Shadbolt, N., Van de Velde, W., & Wielinga B.,

2001. Knowledge Engineering and Management – the

CommonKADS Methodology. The MIT Press,

Cambridge, Massachusetts, London, England.

Tansley, D.S.W., & Hayball, C.C., 1993. Knowledge-

Based Systems Analysis and Design - A KADS

Developer’s Handbook. Prentice Hall, UK.

Wiratunga, N. & Craw, S., 2000. Informed Selection of

Training Examples for Knowledge Refinement,

Proceedings of the EKAW2000, 233—248.

ICEIS 2005 - ARTIFICIAL INTELLIGENCE AND DECISION SUPPORT SYSTEMS