Technology and Differently Abled People


Artificial Intelligence
Arne Svensk

Decision Making Support in Dealing with Violent Behaviour
The SVARNE Expert System

The work of using natural intelligence (see the last chapter) to develop better, more concrete aids for mentally retarded users continues. Artificial intelligence (AI) offers a wide range of possibilities as well. There is much in AI technology that can be put to good use by people with different kinds of disabilities. It is time for the field of (re)habilitation to actively pursue the development of intelligent machines.

AI for diminished IA

"AI" stands for artificial intelligence. "IA" designates intellectual ability. Is it not a challenging thought that AI can be useful in the IA context?

CERTEC has chosen to start this portion of its work with an expert system meant to give support to staff who encounter violent behaviour in caregiving. Here is the background:

During my eight years as director of a residential home for mentally retarded pupils with special needs, I often came in close contact with violent behaviour in the programme. This home was the only one of its kind in the whole county. There were, of course, many reasons as to why a pupil was placed with us, but violence and aggression were often the determining factors.

Many special education schools were poorly equipped to deal with violent situations. This resulted in the pupil being sent to us even though the living arrangements in the local areas worked out well. Each one of us knows that it is not good for a person to lose his whole network of normal contacts when he is already in bad shape. Those who have difficulties expressing themselves in other ways, often resort to violence and aggression in a crisis. This was one of the reasons why it was a common occurrence in our group home for young people.

Violence and aggression, though, are unacceptable means of expression. Friends and staff often showed their rejection of this behaviour by withdrawing. We on the staff were sometimes the only fixed points in the lives of those students who lacked a basic sense of security. If we were frightened, it reinforced their insecurity and lack of self-confidence.

Violence often results in a high turnover of personnel. Caregivers feel inadequate when they are unable to find reasons why the pupil suddenly becomes aggressive. Working weekends is hard for them to manage. They know that they have thirty hours of work before them in which they lack the methods to understand and/or control the violence that can erupt at any moment.

In my experience, violence is the most signigicant reason why caregiving staff decide to quite and do something else.

This was also why CERTEC's development of AI for caregiving personnel started with a decision-making aid to analyse the causes of violent behaviour. In its present form, decision-making support should be seen as a kind of intelligent computerised checklist. But it can already be used as a basis for defining common concepts and search strategies.

Decision Making Support in Dealing with Violent Behaviour

As a new director, I lacked the ability to systematically search for the underlying explanations of violent behaviour. I was even less capable of finding methods that could stop or reduce it. Now that I think about it, it reminds me of the unsystematic way I approach problems with the electrical system of my car.

I know that there are different components in the system and I know the names of some of them such as the distributor, cables, spark plugs, starter and generator. I also know from experience that moisture in the electrical system is a common reason for a car not starting. Another reason is that the battery is dead. I start my search by checking the battery. If that works I continue by removing the moisture from the distributor with the help of a special spray. If the car still does not start, I get a bit desperate and bang on the relay and ignition coil. The last thing I do is to unscrew the spark plugs and wipe them off.

If the car does not start in spite of all these measures, I usually leave it alone for a few hours and hope that time will heal all wounds. Unfortunately, it seldom does. The next step then is to get in touch with the nearest mechanic.

Intellectually, I know that there are systematic search systems which can rule out errors one by one in different parts of the electrical system. This I know from my observations of how experienced mechanics approach the job. Unfortunately, I do not remember what to do for long because I have learnt it mechanically without understanding the underlying principles that guide the search.

When we try to analyse the causes of nervousness or anxiety in our residents, our approach is often as amateurish as my way of searching for the problem in my car. We know which factors often cause anxiety but we lack a systematic search procedure that could rule out some and give weight to others.

Of course, it is not so easy to find just one explanation as to why someone has been agitated or violent. I believe, though, that if better search strategies can be found, we could concentrate our efforts on fewer areas and so render the search more effective.

One way of developing a search system is to let an expert in the area tell what she does to reach the goal. These systems already exist for cars, for example, in the American army. A clever and experienced mechanic has reported what she does to discover certain problems. The information was fed into a computer. Then, when someone else has something wrong with his jeep, and is far from the garage, he can use the computer as an aid in deciding how to fix it. The computer guides him forward with such questions as:

"Are the head lights shining?"
"Yes."
"Then check . . ."
"Does it click when you turn the ignition?"
"Yes."
"Then check . . ."

Personally, I would have nothing against having one of these expert systems in my car.

This leads to the following question: Is this only possible for cars or are there experts on anxiety, agitation and violence? If so, can this knowledge be formalised? I am convinced that the expertise exists and I believe that the knowledge in question can be captured.

I know from my own experience as a programme director, that each year I have become more systematic in my analysis. I have learnt to differentiate the important from the less important. Concerning violence, I began to discern a pattern where one had not existed before.

Though it feels presumptuous to call myself an expert, I still want to share my experiences with others. Why should a new director or caregiver have to make the same mistakes that I made? Is it reasonable that the staff and residents must constantly be exposed to beginner errors? It is considered a given that a flight captain would train hundreds of hours on a simulator before she takes her place in a jumbo jet with 200 passengers. Before the flight, the crew go through a checklist to discover possible problems. I know of no passenger who has been suspicious of the captain because she did not know the checklist by heart.

But when it concerns care for the mentally retarded, we think in a totally different way. For some reason it is not acceptable to refer to checklists for clues and decision making support. Human problems are thought to be far too complex and full of nuances too numerous to be crammed into a few manuals. The only thing considered of value is to spend years going through the mill and in that way gaining some kind of intuitive feel for how to solve problems. I believe that this way of thinking is very dangerous and that this outlook ends up, in the long run, doing more harm than good for those we want to help.

How information is handled in programmes for the retarded

A major problem I had as a director of a residential home for pupils was with the transfer of information. This is an important part of analysing any situation including violent ones. Therefore, I will try to thoroughly examine the steps involved in handling information.

In addition to pupil residential facilities, I have been responsible for three group homes for adults. The table below shows staffing and the number of residents living in the different units.

Pupil Residential facilities vs Group Homes for Adults

A pupil's social network could look like this:

A Pupil's Social Network

The pupil is in contact with about 40 people. Much happens in all the different relationships that young people with emotional difficulties have. So much so that daily communication is needed to and from the home about what is going to happen. The pupil himself is unable to verbally report what has happened for the most part. That job must be left up to others.

Many time, a telephone call means that I or someone else must get in touch with others in or outside the residence. What follows is a good example of this.

This is the result of one simple phone call. On other occasions the chain can be considerably longer. This happens especially when the questions have to be cleared with all the different parties concerned and when the answers affect many others. When violent behaviour occurs the flow of information increases dramatically and, as a result, so does the amount of time it takes to inform each other.

There are at least two aspects to consider in these conversations. The first concerns the actual content, which is that Stina will not be going home this weekend. The other concerns how we inform Stina: we all have to express ourselves in more or less the same way.

If the taxi driver does not know about the cancelled weekend on Thursday, it is very possible that he will say to Stina on Friday morning, "Won't it be fun to go home tonight?" If those working at the residence then spend the whole morning talking about the cancelled trip, Stina will, of course, be unsure about what really is going to happen that evening.

Facts are not all that need to be conveyed. It is just as important that we all have a similar way of explaining what is going to happen. Stina might think that the reason she cannot go home is because she happened to break a vase the last time she was there. It should not be the case that someone casually says, "So you are not allowed to go home this weekend?" Stina would immediately take this as a confirmation that it is her fault her parents do not want to meet her.

All of this means that those who meet Stina on Friday must agree to say, for example, "I heard that your mother is ill. That's too bad. You were so looking forward to going home. But I heard that you will be going home next weekend instead."

For some pupils a certain wording can have a strong, soothing effect while one that is just slightly different can be catastrophic. It is necessary at times to put much effort into formulating what we are going to say and to making sure that everybody says it in about the same manner.

What I am trying to demonstrate with this example is that the amount of information to be conveyed can be considerable even through it concerns only one telephone call and one pupil. When many of these types of phone calls occur every day, it results in the following: someone forgets to pass important information on, or it is not passed on to everybody concerned, or it is not conveyed correctly, or it is simply deemed unimportant.

In the long run

Some daily information needs not only to be shared but to be saved for shorter or longer periods. This means that someone has to sit down and determine what is important and what is not as important. The important information has to then be documented and stored so that it is readily available.

Information can be divided into two categories: what comes in on a daily basis and what is collected over time. For people with mental retardation who are unable to tell their own stories, it is vital that this information is documented. This is the only way to reconstruct events backwards in time and connect them with what is happening now and that at first glance appeared incomprehensible. If you do not know that a pupil was bitten by a dog when he was five years old, it can be difficult to understand why he does not want to go to the park when many people are out walking their dogs. The pyjamas that you are absolutely forbidden to launder can be the only memory left of mum, etc.

My experience is that most of this knowledge is conveyed orally and is not written down. People who have worked a long time at a residential facility find out about these things when they are first employed through what others tell them. They fill in the rest along the way through their own experiences with the residents. When new staff members start they sometimes receive part of the colletive knowledge the first day. Usually, though, they receive it in small portions stretching over a longer period, often in connection with some incident which makes something that happened in the past relevant.

A central problem is that the newly employed staff member has a difficult time actively seeking out the information on his own. He cannot just go to a file and search through it because it is seldom written down. He is totally dependent on somebody who is willing to share these experiences. They are not always willing or able. There can be a number of reasons for this unwillingness. Many caregivers do not really understand just how important their observations can be in giving a clear picture.

There are dangers in depending too much on oral information. This kind of knowledge is physically connected to particular people. If they disappear, important pieces of the puzzle can be lost forever.

Try to imagine how it would feel to lose all your photos from childhood; pictures of your best friend, the first day of school and that little puppy. It would be a great loss and we might try to reconstruct the past by making copies of relatives and friends' photos. We would go to this extent although we could just as well talk to some of them about the past and in that way recall important events.

For those who can neither speak nor write, other means must be used to preserve their memories. One way is to have someone help by writing down important events in a "memory box" that stays with the person when they move or can be used to tell about themselves to new friends.

All in all, written, pictorial or by-other-means recorded information is extremely important in the care for the mentally retarded.

Violence and lack of information

Violent behaviour accentuates the difficulties that result from lack of information. Personal sensitivity comes into play as well. These constitute some of the limitations of the oral tradition. Those people who are closest to the resident are the ones who are the most engaged in supporting her. This means that all their physical and mental energy are directed towards reducing the aggressive behaviour. There is not much room left for conveying or seeking information.

It is also in connection with violence that much of the collected knowledge disappears. At that point the staff often quit. Since they give only one month's notice and it takes about six weeks to employ someone new, the person who is quitting does not have the time to personally inform his replacement. This means that those who are left and pressed for time must try to get as much information as possible from the experienced caregiver to be able to repeat it to the new employee.

It has been my experience as a director that if this kind of situation is not handled with the utmost care, you end up in a vicious cycle in which the total amount of violence increases.

Characteristics of violent and aggressive situations

We hope that this background description will clarify why we think it is so vital to design a means of giving decision-making support in violent and aggressive situations. In the list that follows, I have summarised the arguments in favour of a more systematic means of seeking knowledge in this area.

Why systematize the search for knowledge?

Those who resort to violence generally do not feel good about themselves and need help quickly.
Violent behaviour scares both friends and staff. They have difficulty coping or refuse to stay on in the residence.
Violence increases the need for information both internally and externally.
When violent and aggressive behaviour occurs, everyone is so occupied with dealing practically with the situation, that no one has the time or makes the effort to search for significant factors.
Those who are most interested in reducing violence often lack the systematic search strategies necessary for quickly finding possible causes.
Those who have the theoretical framework are usually not as strongly motivated to do something about violence.
Habilitation efforts are negatively influenced if staff members are afraid.
Violent behaviour in group homes has far reaching consequences.

The SVARNE Expert System

CERTEC began constructing an expert system with the working name of "SVARNE" in l993. (The SVARNE Expert System in its present form is meant to be solely an educational system. Advice from the county attorney has been sought and will continue to be sought to insure that SVARNE’s future use in concrete situations will not violate existing legislation.) We chose to build the whole system from the ground up to achieve a better understanding of how it works. The team has consisted of experts in the actual subject area (myself, Anre Svensk), a knowledge-engineer (Charlotte Magnusson) and a programmer (Ola Liljedahl). We have also had access to an expert system's specialist (Jan Eric Larsson), who during the initial phase of the project worked as the project manager.

The first prototype system has been built and runs on a PC. In this system you can find some of my expert knowledge concerning violent situations in the care of the mentally retarded. How have we approached the analysis and structuring of what I think I know?

The first question the group presented to me was: "Can you tell us how you usually think when a violent situation arises?" It did not seem too difficult a question to answer but it turned out that it was hard for me to tell others how I went about it. Even more troublesome was the discovery that I could not give a satisfactory explanation to myself. I had to sit down and carefully go through all the violent situations I had experienced to try to find what kind of search strategies I usually used.

Finally I found a basic pattern that looked like this:

1 Do I recognize this violent situation? If the answer is "Yes", we do what we usually do. If the answer is "No", go to point 2.

2 Has anything special happened?

When, after careful consideration, I dared to write these two points down on paper, I began to question the whole project. Is this what it looks like? Is this all of my "expert knowledge"? Does my search strategy for violence and aggression consist of only two insignificant points - after ten years as a supervisor?! How miserable, to say the least!

But things started to happen after I had presented these points to my co-workers in the project. They helped me by asking question that led to follow-up questions, and gradually it seemed that my strategies were a little more advanced than these two points after all.

We started drawing a tree. The branches were made up of questions that I usually ask myself when I try to find causes of violent behaviour. At first all the branches grew out of the same spot on the tree and were about equally thick. After a while some branches got thicker and from these, smaller branches grew. In other words, the search system became more hierarchical.

SVARNE TREE

Schematic drawing. First, the violent behaviour is divided into three categories: violence towards oneself (most serious), violence towards others and violence towards objects (least serious).



The picture above is a schematic drawing of the expert system's current design.

The program starts by asking for resident identification. When the answer is given, a square containing facts about the individual appears. Besides vital statistics, it lists other important factors such as if the person has epilepsy or allergies. The number of situations in which the person has reacted with violence and aggression are also listed. These are described in short sentences such as, "Can become very angry if a button falls off his jacket or if he stains his clothes."

After the factual square the expert system kicks in with the three subcategories of violence: towards oneself, towards others and towards things. I rank them according to degree of seriousness: violence towards oneself is most serious while violence towards objects is least. The expert system asks the user such questions as: "Is it true that the violence is directed towards the person himself?" Depending on the answer, different branches of the tree are followed and suggestions are given for possible measures to be taken.

Violence and aggression can have physiological explanations and so you could get the following kind of question: "Does he suffer from migraines?" If you are uncertain of how you can objectively observe a migraine, you can ask for a list of observable behaviours that can assist you in answering the question.

Work in progress

Even though the expert system is not completed, the program can still be used. It has already proven its value by increasing stringency and in that way the usefullness of what I think I know.

I question whether there are any other means of helping me gain better insight into my experientially based knowledge to the same high level that an expert system can. It feels as though I have been able to considerably sharpen the clearness of my thoughts in one leap.

Advantages and disadvantages of using an expert system in caregiving worth noting:

Conceivable expert system advantages

The advantage of using checklists on computers is that they can be done individually. The information on the list is specific to a particular person. The checklist forces the user to think through factors that otherwise might be considered uninteresting. This relieves the memory load.
Expert systems demand that one states concepts and methods clearly. You have to be specify what you mean by different expressions. This results in all who work in the residence having a common base to start from.
By expressing thoughts in plain, clear language, they are easy for all to understand and use as topics for discussion.
The system is educational. New employees can benefit from the knowledge that an experienced care provider or the relatives have accumulated over a long period of time.
An expert system can continually be improved through the addition of new information. We can gain from the knowledge of others and do not have to start all over.
Important information is collected in one place. This is much easier than having to look through different files or talk to different people.
Information is at hand for those who are most motivated to seek it out and use it.

Conceivable expert system disadvantages

The gathering of so much information about a person in such an accessible format can be seen as an invasion of privacy.
Instead of using the program as a support and intellectual reinforcer, the staff instead interpret the expert system's statements as truth.
That the system gives such elementary decision-making suggestions that it is considered useless.
That it is perceived as being so difficult to add new information that it cannot be depended on to be up-to-date.
The system forces one to formalise knowledge that it is really impossible to formalise.
Fewer contacts with specialists who have approaches that differ from the expert system's.


Preface
Table of Contents
1. Introduction
2. Natural Intelligence
3. Artificial Intelligence
4. Isaac