Certec Report 4:1998

Expert Systems and Learning

Charlotte Magnusson, Certec

Introduction

Since 1993, Certec — Center for Rehabilitation Engineering Research at Lund University has been working on the development of knowledge-based systems, so-called expert systems (see Appendix 1), for various contexts related to rehabilitation and the care services. Our work has resulted in the following expert systems:

 

Svarne A decision support program to help care services staff analyze the causes of violent behavior
SRF Expert A cane training program for people with acquired visual impairment
Ask Freya A system for structuring documentation about Freya, a woman with autism
The Math Expert A system designed to help students analyze their understanding of the concept of percentages
Question Box A program for transferring knowledge about a student in a special school
Socrates A program that helps the user create and test philosophical definitions

Furthermore, work is ongoing in the projects Compulsive Thoughts, SeeIT, and Knowledge-based systems - especially for people with autism.

In each of these projects a number of results have emerged –not least that the knowledge-based system is an important result in itself. While working on a growing number of such systems, we have become increasingly aware of the fact that building knowledge-based systems is an extraordinarily effective way of developing one’s own thinking. Consequently, it is only natural that we have begun to look for ways in which this technology could facilitate learning in a more general context.

We are thus moving away from research projects in which we build various knowledge-based systems in order to focus more on the interaction between the technology and the individual in the learning situation which the creation of a knowledge-based system involves.

Nevertheless, the many completed projects have, of course, yielded a large amount of experience concerning expert systems and learning. This report is an attempt to sum up this experience and to place it in its new context.

 

Expert Systems at Certec

In 1993, Certec began the construction of an expert system, Svarne, which was intended to be used by care services staff as a decision-making support program for analyzing the causes of violent behavior. The project has been extensively documented in the reports Svarne, beslutsstöd vid våldssituationer (Decision Support for Vi knowledge olent Situations) [1] and Kunskap på burk (Canned knowledge) [2].

Once Svarne was in place and could be demonstrated as an example of how an expert system can be of use, new project ideas gradually emerged. In 1995, Synskadades Riksförbund (The Swedish Organization of the Blind and Partially Sighted) contacted Certec indicating an interest in creating an expert system about cane training for individuals with acquired visual impairment. A project was set up, leading to the creation of a prototype in which the major part of the knowledge concerns basic cane technique and mobility. In our contacts with Freya [3],[4],[5], it soon became obvious that an expert system might be a means for achieving structured documentation of significant situations. The expert system Fråga Freja del 1, frågor och fragment (Ask Freya part 1, Questions and Fragments) was developed. (There is also a program called Fråga Freja del 2, Freja gör framsteg (Ask Freya part 2. Freya Makes Progress.) The Matteexperten (Math Expert) expert system was created to provide answers to the question why deaf children on average perform less well than hearing children in solving math problems, [6],[7] (the project started in 1996, in collaboration with Elsa Foisack, Östervångsskolan, Lund). Agnetha Dyberg-Ek, a teacher at the special school in Hörby, contacted Certec in 1997 with a request for the development of an expert system to help a student with autism change schools. This resulted in the creation of the FrågeLådan (Question Box) expert system prototype. Moreover, inside Certec, there were ideas concerning expert systems, information, and knowledge generated by Bodil Jönsson’s question "what is the difference between information and knowledge? ". This led to the creation of the Socrates expert system in the autumn of 1997 [8].

 

Svarne

Svarne was Certec’s first expert system project. It was a large project which produced many results that were confirmed in subsequent projects and therefore it warrants a more detailed description.

The intention was to find out whether it would be possible to handle in an expert system the knowledge of a person (Arne Svensk) with many years’ experience of working in the care services. It proved possible, and the results of our work on Svarne included a method for managing this type of knowledge. The knowledge contained in Svarne is limited to violent situations.

A Java version of the decision support part of Svarne can be found at http://www.arkiv.certec.lth.se/expertsystem/expertsystem/JavaExpert/demo.html.

svarne_träd.jpg (44902 bytes)

A schematic picture of the decision-making tree in Svarne. First, the incident is characterized as "self-destructiveness", "violence against others" or "violence against objects." The "violence against others" category is divided into the subcategories "person-related violence," "situation-related violence," and "independent violence". Person-related violence means that there seems to be a connection between one or more individuals and the incidents in question, while situation-related violence seems to occur in certain specific situations. Independent violence is the term used when it is difficult to see any particular connection with respect to time or place. Finally, at the bottom under each heading, a division has been made into "physical reasons," "social reasons," and "psychological reasons".

Building Svarne

It turned out that the expert system itself was an excellent tool for setting new thoughts in motion and extracting new knowledge. The new knowledge could then be added to the program and in turn generate further knowledge, and so on. Consequently, our work was based on successive prototypes of Svarne. We also found that it is advantageous to work with only a few experts when putting together the overall structure, while it may be better to use a large group when filling the system with practical advice and experience. Moreover, we would never have been able to develop Svarne without collaborating closely with several locations.

The method we used for building Svarne can be summed up as follows:

 

    Use the expert system as a tool for extracting knowledge. Involve future users in the development work. In order to ensure the usability of the system it is important to develop it in close proximity to the reality in which it will be used (preferably in it, if possible).

    Begin working with one expert. Be careful in selecting this expert. It is not only the knowledge the expert possesses that is important, but also how the future users perceive him or her. It is a definite advantage if the expert comes from inside the user group.

    When a first functioning prototype has been developed – show it to other experts. If they think that what has been done so far seems reasonable, it is time to expand the expert group. However, be careful not to make the group too large, two to three people should be sufficient. At this stage, the focus of the work should be on creating an overall structure for the knowledge.

    When a good overall structure has been created, it is time to contact a large group of experts who will help fill the structure with examples, advice, and experience. This can be done at a one-day seminar, for example, where all the experts use a computer to go through the system simultaneously and discuss it.

An important person in this context is the so-called knowledge engineer (in the Svarne project mainly, Charlotte Magnusson). The knowledge engineer prepares the knowledge and enters it into the system. In Svarne, we found that the meeting between the knowledge engineer, with a background in science and mathematics, and the highly experienced care services staff, who participated in the project as experts, was very important to the final result. Another important meeting was the one between technology and people. Technology, in this case an expert system, forces the experts to be more precise than they usually are, making the knowledge in the expert system more clearly worded and better structured. Subsequently, when the expert sees his or her knowledge in motion in the expert system, he/she may have "aha-experiences" that would be difficult to attain in other contexts.

In the words of Arne Svensk:
"I doubt very much that there is any other method that could help me gain insight into the knowledge my experience represents to the same extent as the expert system. I feel that I have succeeded in sharpening my thoughts in a single leap."

 

Using Svarne

Svarne is based on a division of labor between the computer and the user. The computer keeps track of the overall structure and provides clues and thought-support, while the user contributes his or her experience and intuition. This requires a great deal of the user as well as of the system. Svarne is not designed to be used directly in an emergency situation. The system may be of use before something happens, by providing information or by being used for training purposes, and afterwards to help in giving structure to what has happened.

It is important to develop good routines for how to use a system like Svarne. With hindsight, we can see that we should have stressed this point more than we did. To a large extent, we left it up to the users to establish these routines, with good results in some cases, but not so good in others. It probably would have been better if we had helped the users by providing more concrete examples of how the system can be used. On the other hand, it is difficult to know this before users have tried the system for a period of time. One simply cannot know until one has tried. After all, we now have a number of examples of good ways of using Svarne.

 

Thought-support / checklist

Perhaps the most obvious use for Svarne is as thought-support or a checklist. By running Svarne, the user is provided with reminders and ideas, as well as practical advice. In a wider sense, the structure contained in Svarne can help the experienced user to better organize his knowledge. When used in this way, the system can also counteract "cognitive tunnel vision" (focusing on one possible solution to the exclusion of alternative or complementary solutions), which is a common affliction in difficult situations. Another well-known tendency is that we often think that other people act the way they do because of their personality (while we prefer to explain our own actions by referring to environmental factors). When used as a checklist, Svarne can counteract this tendency by reminding us of a number of important external factors.

 

Knowledge-extraction Tool

This way of using Svarne has been described above. It involves using the expert system to extract new knowledge about a specific problem area.

 

Education

Demonstrating Svarne during a lecture while discussing the program is a superior method of activating an audience. In this context, the program acts as a communication tool and it is an excellent way of setting knowledge in motion. Naturally, this presupposes that the audience possesses a certain level of knowledge and experience. When we demonstrate Svarne to people who are not care professionals, we do not get the same response at all. The only exception is when we demonstrate the system to groups with similar problems, such as staff in the psychiatric care services. Incidentally, this also emphasizes how important it is that the user has a certain amount of knowledge – the knowledge contained in the computer is not enough.

 

Introduction of New Staff

New staff are often introduced to their workplace by being shown around by an experienced staff member, who tells the newcomer about policy and other things that are important to know. This method of introduction is not always efficient. Firstly, important information may be forgotten and, secondly, a lot may be said about relatively unimportant matters. It is also quite likely that in this kind of introduction the focus will be on problems, not on positive aspects. Using Svarne when introducing new staff may ensure a better introduction.

Tony Persson, who was working in a group home in Södra Munkarp during the greater part of the Svarne project, began using Svarne when introducing new staff. He would begin the introduction by telling the newcomer a little about Svarne. Then, they would go through the program together, looking at all the alternatives and discussing them. Next, the new employee would sit down and try out the program for a while. Many of those who were introduced to the group home in this way thought it was a very good method, allowing them to try out the program even if Tony was not present at the time. On one occasion Tony and a new staff member used the program after an incident that had left the newcomer very upset. By having the opportunity to sit down at the computer in peace and quiet they were able to work out what they believed had gone wrong.

Used in this way, the computer keeps track of the agenda while the deep knowledge comes from the individual with the experience. Moreover, the expert system allows a new employee to actively search for knowledge. The program has a table of contents from which the user can proceed by asking questions. Without a computer there is no such basis from which to move ahead. All the knowledge is contained inside the heads of more experienced colleagues, which means that a new employee is obliged to wait for them to tell what they know. Furthermore, with Svarne there is no risk of missing important points.

 

Documentation

Using Svarne staff are able to document important information about individuals. This facility was used in all the locations which participated in the project. However, the information was not always kept up-to-date. It is important to develop routines for ensuring that the information in the system is accurate. In the Oscarshem group home in Lund, a summary is written at the end of each month based on the daily entries, which are then discarded. In such a case, it may be suitable to establish a routine whereby relevant information is entered into Svarne at the same time. Making staff members responsible for checking the information in Svarne, updating it at regular intervals if necessary, is another possibility.

Tony Persson also noted that it is usually better if two people collaborate when entering the information into Svarne. In this way, they are able to help each other to put their thoughts into words, and it also makes using Svarne more interesting.

 

If you are sitting by yourself in front of the computer it does not take very long before you have gone through all the information, and after a while you may get bored because nothing new is happening. When two of you are working together new stories and new information keep emerging, which quite simply makes the system more fun to use.

One problem is that there is often no immediate reward for the work staff members to put into Svarne. The users themselves must discover how the system can be of use to them; otherwise they will not be motivated to keep the information up-to-date. If this does not happen because they have decided beforehand that other methods are better, they will not be very motivated to input the information. Nevertheless, we have found that the very existence of Svarne has an influence on routines and working methods. This is true not just with respect to thoughts about violence, but also with regard to the importance of documentation. For example, purely oral introductions have been supplemented with a written checklist. The checklist in Svarne comprises 110 items, but each of these is of a fairly general nature. It is possible to incorporate custom rules to create a custom checklist, but users have found this difficult. The reason for this is not that the program is difficult to use, but rather that structuring one’s thoughts has proven the big stumbling block.

A common argument against Svarne and computers is that there is not enough time. In reality, lack of time is a strong argument in favor of gathering information and knowledge in a system like Svarne. If maintained properly, this may eventually result in less time being spent on finding information or transferring knowledge.

 

Summary of Svarne

Svarne has shown that it is possible to use an expert system to transfer knowledge within the care services. It has proven an effective educational tool by setting thoughts in motion and making users think along new lines to some extent. Svarne has also highlighted the need for documentation, and it has lead to the development of many new projects.

So far, Svarne has had a very favorable, indeed enthusiastic, reception from all the workgroups we have met. There are probably several reasons for this, but the following may be some important points:

    first, we introduced Svarne and the ideas behind it to the entire workgroup

    we did not focus on the technology, but rather on what it can do

    we focused on a limited area

    Arne Svensk, who has extensive experience of the care services, participated in the project from the very beginning.

    we dealt with a difficult problem in a serious way - "finally, someone is listening to us," as some of the participants expressed it.

Perhaps one of the most important results so far is that according to those who have taken part in the project it has prompted them to begin looking for new reasons for violence - primarily, physical pain as a reason for violence. The fact that 10% of the general population have been diagnosed as suffering from migraine indicates that not enough attention is paid to pain as a cause of violence. During the entire project we did not meet a single care professional who had even heard of a developmentally disabled person with that diagnosis. In addition, we have seen examples of how long it takes before toothaches, ear infections, urinary tract infections, and even open wounds are discovered.

At the same time, we have experienced the importance of the social environment when developing and introducing new technology. We have had problems because users were unfamiliar with computers, but also because Svarne requires staff members to change their routines (perhaps even the entire organization). These and other environmental factors may explain why Svarne worked best as an educational tool. The fact of the matter is that it is easier to influence an individual than to change an organization (but, naturally, we are hoping that the individuals we have influenced will help develop their organization in the long term).

 

SRF Expert

One of the spin-off effects of the Svarne project was that SRF (Synskadades Riksförbund — The Swedish Organization of the Blind and Partially Sighted) contacted us indicating an interest in creating an expert system concerning cane training for individuals with acquired visual impairment. The project has resulted in a prototype in which the major part of the knowledge concerns basic cane technique and mobility. A report describing the prototype has been presented to The Swedish National Board of Health and Welfare.

SRF Expert allows the user to enter information about diagnosis, visual acuity, field of vision, sensitivity to light, scotopic vision, etc. Moreover, the system always displays the following text: "There is a myth about blind people, according to which only certain, very gifted, blind individuals are able to become highly proficient at getting around. The myth is false, but hard to dispel. Through hard work and a belief in their own ability, all individuals with a serious visual impairment but no other disabilities can become very proficient at getting around in their community with the aid of a cane. The perpetuation of this myth results in fewer rehabilitation opportunities for people with visual impairment!"

SRF Expert guides the user through a number of questions intended both as reminders and as throught-support.

SRF_expert.jpg (12892 bytes)

A screen in SRF Expert. Like Svarne, SRF Expert ask questions beginning with "is it true that... ". In this case, the text reads as follows: "Is it true that Peter is holding the cane in the standard way (in his strongest arm with his index finger resting on the cane?)"

The program offers advice based on the answers given by the user. The advice is saved and can be printed out later. In its present form the expert system contains 60 rules and suggestions for action (exercises and explanations). The program was tested at a couple of vision centers in the Stockholm area in August 1997. Interested readers may contact Gunnar Sandström or Bertil Sköld at SRF for further details.

 

Fråga Freja (Ask Freya)

FF1.jpg (59269 bytes)

A screen in Ask Freya part 1.

Fråga Freja (Ask Freya) forms part of our efforts to make Freya more visible, both to herself and to those around her. The objective is to enable her to move away from a description of herself as a human enigma and to become an individual whose reactions and wishes are understandable and manageable, particularly in view of her past history.

The following questions form the basis of Ask Freya

    What has happened?

    What does Freya want to happen?

    What do I /those around her think should happen?

 

Certec’s fundamental approach to trying to understand a person whose way of thinking and pattern of behavior appear aberrant is to first ask the person herself. If this is not possible, one must try to bring together various representatives of the world around her to attempt to understand her better. When this has been accomplished, signals from the person concerned, in this case Freya, should be relied on.

 

In Fråga Freja del 1, Frågor och Fragment (Ask Freya part 1, Questions and Fragments) we have succeeded in making visible parts of a know-how, a know-why, a know-what and a know-when that Freya and those around her agree on. The program also includes features for easy and structured documentation of events.

The design of a good Ask Freya system is not obvious. These are some of the possibilities:

    example-based

    rule-based

    a game presenting various scenarios, where the questions may concern both causes and appropriate action. Naturally, Freya will win the most points in this game.

    a critiquing system where the user must decide how to act in a specific situation. The system provides feedback by commenting on and critiquing the measures suggested by the user and by suggesting alternative action.

Unfortunately, Freya’s situation has changed once again, with the result that our work on Ask Freya has been suspended for the time being.

 

MatteExperten (The Math Expert)

MatteExp_intro.jpg (59366 bytes)

A screen from The Math Expert

Studies [6], [7] show that on average deaf children perform less well in solving math problems than hearing children do. Despite the fact that, in principle, the ability to hear is not necessary to solving mathematical problems, something interferes with the learning process in deaf children. Some students in the higher grades lack knowledge of basic concepts such as decimal numbers. There are probably several concurrent reasons for this, but it has not yet been established exactly where the problem lies. The fact that textbooks for deaf students are almost exclusively available in Swedish, their second language after sign language, is probably an important factor.

MatteExp_video.jpg (13014 bytes)

A scene from a video sequence in The Math Expert. A considerable amount of work goes into producing each task in sign language. A great deal of care must be taken in preparing both the content and the wording.

To provide an alternative, we developed an expert system prototype in collaboration with Elsa Foisack, a research student and teacher at Östervång School in Lund (in which we decided to include calculation of percentages only). The system can formulate mathematical problems in various ways by means of pictures, video clips and text. It can then help analyse the user’s knowledge profile on the basis of the answers he provides.

MatteExp_fråga.jpg (55008 bytes)

A question in The Math Expert. A question may comprise text, an image or a video. Answers may be written in a textbox, or be selected by clicking on the correct alternative.

The program is intended as a tool to help the student see her own knowledge profile in the field of mathematics: she can find out what she knows and what she needs to learn (and no longer be stuck in "I don’t understand anything"):

Example No. 1: "I don’t understand percentages, but I understand what a half and a quarter is."

Example No. 2: "I don’t understand 20%, but I understand 50%."

Example No. 3: "I have a good understanding of percentages if the information is presented to me in sign language."

Since it is important that a program such as this be expandable, the project also includes the development of a simple editor, Hotspot, connected to the expert system. To date, prototypes of the editor and the program have been developed. In addition, we have developed a simple training program called Math Challenger, using the same type of questions as the Math Expert.

 

Frågelådan (Question Box)

This project was created to respond to the needs of a small boy, whom we will call Peter, who has been diagnosed with autism and a developmental disability. Peter lives in a foster home. When he started school, he was described as a person with many difficulties, such as refusing to eat, sleeping difficulties, sound and machine fixation. He was also extremely dependent on having an adult close by.

When the project started, Peter had spent six years in a special school. He has been in a small class with four students who all need the same type of structure. In close collaboration with his family, the school has given Peter a sense of security through clarity, structure, and visual diagrams. He is now able to work more or less independently in areas that interest him. He has an aptitude for domestic work and exploring nature, and he shows great endurance. He likes creative work, such as painting and music. Over the years, Peter’s self-destructive behavior has become increasingly manifest. It is a means of exercising power when he feels insecure because of a lack of visual structure.

The need for documentation about Peter was brought to the fore when he was moving to a new special school. He would be in a new environment with new staff, and there was a need to transfer information and knowledge about him. Video documentation as well as written documentation was available, but in the spring of 1997 it became obvious that this was not sufficient for dealing with Peter’s self destructive behaviour.

This was the starting-point for a first prototype of the Question Box program.

FrågeLådan.jpg (19827 bytes)

A screen from Question Box. The writing on the three buttons says Select, Zoom, and Interrupt. The coffeemaker is a substitute for a picture of Peter’s home.

It was based on the idea that it would be possible to impart knowledge about Peter and the practical solutions that had been worked out over the years. Question Box enables you sit down with Peter in front of the computer to look for the solution to a destructive situation. For this purpose, the program contains pictures for Peter to react to, as well as text to help those around Peter understand him better and thereby find solutions to difficult situations.

We had a vision that, with the aid of this system, Peter would be able to guide people around him to solutions to problems that might occur or to give them the codes that would enable them to move forward. Unfortunately, various practical (organizational) problems have prevented us from continuing our work with Peter. However, we have been introduced to another student in the special school system who might benefit from a similar system.

 

Socrates

Socrates is a program that is intended to provide help in establishing definitions. It assists users in bringing out significant elements, necessary and sufficient conditions, and all other elements required to produce an accurate and tenable definition. The program also comprises a number of tests which help the user find flaws in his thinking. Socrates was developed in collaboration with Jan Eric Larsson of the department of information technology and Bertil Rolf of the department of philosophy at Lund University.

 

Compulsive Thoughts

We are collaborating with Jan Eric Larsson of the department of information technology at Lund University and Jan Erik Nawrin, a psychiatrist with the Skåne County Council, on a project dealing with obsessions. The existing program asks a number of questions and saves the answers, allowing the user to narrow down a problem area and to monitor the patient’s progress.

The project started out as an expert system. It was based on the idea of using the answers provided by the patient to devise a treatment tailored to his needs. Early on in the project, however, we realized that we must begin by investigating what the relevant questions are and how different people answer these questions. We have not abandoned the idea of an expert system, but at present the project is in an investigative phase.

 

New Projects

Recently, we began two additional projects related to the field of knowledge-based systems and learning. One is called SeeIT and is being carried out in collaboration with Jörgen Gustafsson of Syncentralen at the Ryhov County Hospital in Jönköping. The second project is "Knowledge-based systems - especially for people with autism" and is being carried out jointly by Certec and Agnetha Dyberg-Ek, a teacher at the special school in Hörby, and Abulkadar Faarax, Simon Nyman, Stefan Nyman and Gertrud Grahn-Nyman [ 8] .

 

What We Have Learned

Early in the Svarne project we discovered that an expert system can be a very effective tool for activating, highlighting, collecting, and structuring knowledge. The expert system becomes a tool for its own creation. The only instance where this is not evident is when the first expert has already got such a clear picture of the structure in his mind that further refinement is not meaningful. That, however, is a fairly uncommon situation. Usually the expert believes that she has a pretty good idea of what the structure looks like, and she only gradually discovers the numerous gaps, illogical thinking, and sheer fallacies. All of these deficiencies come to the surface when an expert system is being built and can, as a result, be dealt with.

The other side of the same coin is that an expert system can be a powerful educational tool. And that an expert system provides clearly arranged documentation. This visibility means that it is important to consider how the system visualizes the knowledge it contains. Usually, it is not enough to show the result of a knowledge structure by answering questions and drawing conclusions. Some kind of comprehensive overview is desirable.

We have also learned a great deal about the importance of various kinds of interaction: the interaction between individuals from different backgrounds (between the knowledge engineer and the expert, for example), and the interaction between the individual and the technology. The interaction between several individuals and the technology is especially important.

However, it is difficult to build knowledge structures. It requires motivation and hard work on the part of the user. Naturally, the system must provide as much assistance as possible. But when all is said and done, the quality of the system depends on the quality of the knowledge that has been fed into it. As in so many other cases, GIGO applies here as well: garbage in garbage out. The only difference is that it becomes all the more obvious in the context of an expert system.

If the expert system is to work in the long term it must be easy to expand and change. And it must be based on the user’s own words and pictures.

It is difficult to say which project generated a specific insight. For example, in Svarne we managed from the start to get several elements right. This meant that we did not become aware of their significance until they were missing in later projects. The importance of a comprehensive overview is an example of this. Svarne displays a schematic picture of its decision-making tree, which many of the subsequent programs did not.

The following is a summary of the lessons we have learned from our projects:

 

22 Important Points

    The expert system can itself become a tool for bringing forth and structuring the knowledge upon which it will be based.

    The expert system can make knowledge visible and set knowledge in motion.

    Using an expert system may allow one’s thinking to go deeper.

    An expert system can be an educational tool.

    An expert system can be a form of documentation.

    The individual who runs an expert system gets to follow a train of thought.

    An expert system can function as memory support/thought support.

    The meeting between the individual and technology can be very fruitful and, in this context, it can be an advantage that the technology is distinctly "non-human".

    The system should be based on a division of labour between the individual and the machine. Each should do what he does best.

    The interaction with the knowledge engineer is important and rewarding.

    The development of an expert system puts considerable demands on the participating experts.

    The development of an expert system puts considerable demands on the participating experts’ work environment. It must provide sufficient space for participation in the development work.

    It is difficult for an ordinary user to build a knowledge structure on his own.

    One cannot know if an expert system is feasible until one has tried.

    One cannot predict the quality of the finished expert system on the basis of the first prototype. A meager first prototype may eventually turn into a rich/detailed first-rate expert system. And a fairly detailed first attempt may not get much further.

    The finished expert system will be no better than the knowledge upon which it is based.

    It may be better to be two at the computer rather than one.

    It is important to work closely with future users.

    A visual (or other) overview is desirable.

    The system should be flexible and easy to change.

    It must be possible to use the user’s own pictures in the system.

    The expert system should focus on a limited problem area. This is an obvious and well-known requirement, which applies to all types of expert systems, not just the ones described above.

     

 

A Walking Stick?

In almost all of these projects, both we and the experts involved noticed how working on an expert system developed our thinking. This gave us the idea that it would be interesting to study the interaction between the individual and the expert system technology from a learning perspective. We wanted to investigate whether and how building an expert system may constitute a new gateway to learning and communication. Not just for specially trained programmers but for anybody, and perhaps especially for people with autism.

In order to make this possible, a program must be available where it is easy to create or edit the knowledge structure that will form the basis of a knowledge-based system. Software for editing and changing input is usually called an editor, whereas knowledge-based systems are known as expert systems. Consequently, I have chosen to call this program an "expert system editor".

An expert system editor thus enables a user to create and manipulate symbols, while also permitting editing of relations between symbols and of the rules that apply to the manipulation of these symbols. It must also be capable of displaying the result of the knowledge by means of the knowledge-based system to which a certain symbol and rule structure corresponds. In this way, a special interaction can be achieved.

The different parts of the external system (the expert system editor) act simultaneously –thereby triggering simultaneous internal action in the individual. And vice versa.

The external system can be easily varied – thereby triggering internal variation. And vice versa.

The external system can influence internal attention – and internal attention can be canalised to the external system.

Thus, the external and the internal go together in a way never previously investigated.

An alternative term for the expert system editor might be to call it a walking stick. With it, you are able to do something continuously variable that accompanies you. It works with you and through you, and both parties feel "simultaneity", "variation" and "attention".

 

Problematic Points

Although the 22 points and the expert system editor derive from the same source, a number of the points may, at first glance, seem to conflict with the idea of the editor. In particular, point 10 "The interaction with the knowledge engineer is important and rewarding" and point 13 "It is difficult for an ordinary user to build a knowledge structure on his own". The apparent conflict is resolved by expanding the context from envisaging one user to envisaging many users.

Other points that may appear problematic, at least to those who envision a shortcut rather than a new gateway to learning, are points 11, 12, and 16. Our work on the expert system editor does not afford any "free knowledge". But precisely for that reason it may prove extremely rewarding.

 

The Expert System Editor

A program for developing knowledge-based systems is commercially available. But in order to gain complete control over design, we chose to create our own program.

A first prototype of this expert system editor has been developed. It allows the user to input sequences and build hierarchies and a network of significant elements. The user can save his work and test it by running the knowledge-based system corresponding to a particular structure. To gain an overview, he can either zoom or scroll. Copying/pasting elements is also possible, as is printing.

ExpertEdit.jpg (23777 bytes)

The three ways of building structures available in our present expert system editor. Note that it is possible to input elements to the network structure (right) without connecting them to each other. Accordingly, one can start by inputting the elements one believes should be included and then trying out different ways of connecting them.

Preliminary user tests are ongoing and will probably result in changes. Nevertheless, they have shown that even this early version is usable in many respects and that we are on the right track.

 

References

[1] Expertsystemet Svarne, beslutsstöd vid våldssituationer, Charlotte Magnusson, Arne Svensk, CERTEC 4:93

[2] Kunskap på burk - SVARNE, slutrapport, Charlotte Magnusson, Arne Svensk, CERTEC 3:97,
http://www.arkiv.certec.lth.se/dok/kunskappa/

[3] Free Freya - part 1, Eve Mandre, Bodil Jönsson
http://www.arkiv.certec.lth.se/doc/freefreya1/

[4] Free Freya - part 2, Eve Mandre, Bodil Jönsson, http://www.arkiv.certec.lth.se/doc/freefreya2/

[5] Free Freya - part 3, Eve Mandre
http://www.arkiv.certec.lth.se/doc/freefreya3/

[6] Döva barns utveckling i ett tidsperspektiv. Kunskapsnivå och sociala processer,K.Heiling, Stockholm: Almqvist & Wiksell International, 1993.

[7] Fysiskt handikappade elever i grundskola, specialskola och gymnasieskola - en probleminventering som underlag för ett utvärderingsprogram, M. Myrberg, Stockholm: Skolverket, 1993.

[8] Multimodal knowledge based systems – especially for persons with autism, Charlotte Magnusson, The second Swedish Symposium on Multimodal Communication, Lund, 1998, http://lucs.fil.lu.se/Multimodal/Abstracts/Magnusson.html

 

Appendix 1:

A Knowledge-based System?

Knowledge is a difficult concept. What is it really? Since this report is not intended as a contribution to the debate over the real nature of knowledge, I use the word knowledge in a very wide sense. Essentially knowledge is something that allows an actor to perform an action or a thought. However, some quality requirements must be met if this something is to be called knowledge. The act that is the result of the knowledge must be carried out in a satisfactory manner and, similarly, the statement resulting from it must be true. In addition, the quality of the performance must not be due to a lucky coincidence. If I happen to guess that 2+2 makes 4, it is due to pure luck and not because I possess that knowledge.

The actor who possesses the knowledge can be an individual, but it is not a requirement. The actor can also be a group of people, an individual and a machine, etc. According to this definition knowledge can be both inside people’s heads and in the world at large.

Obviously, a knowledge-based system or program is a program based on knowledge. The difficulty of my wide definition of knowledge is that it means that there are hardly any programs that are not knowledge-based. Instead, it becomes a matter of approach. In a knowledge-based system the focus is on the actual knowledge. It is not certain that the knowledge will become an expert system - perhaps it will be described in a book instead. However, it is often advantageous to represent the knowledge in an expert system since, unlike a book, the former can actually accomplish something on its own.

An important feature when building an expert system is that there are different roles that must be filled. The first is that of the expert and the second is that of the knowledge engineer. The expert is simply a person with sufficient knowledge of a certain field to enable the creation of a knowledge-based system, an expert system, based on this knowledge. The knowledge engineer has the task of processing, structuring, and formalizing the knowledge to give it a form that can be represented in a computer. Possibly, there is a third role involving actually inputting the knowledge to the computer.

In a way one could say that the expert system itself has a role in the process. Thus, there are actually a total of four roles to fill. The same person may fill the three human roles, but the expert system is a computer program and is therefore necessarily located in a computer. One definition of a knowledge-based system could be that all of these roles have contributed to the process resulting in the creation of the knowledge-based system (the expert system).