Figure 14.1a ARTS, the smallest expert system in Holland, based on Bayes statistics (Demonstration version, 1999; with animation).
This chapter discusses:
It is generally known that expert systems and artificial intelligence seem to have become apt aids for researchers and developers. We try to prove that these systems can also become interesting for educational purposes. For an expert system can be adapted for use in education. A pupil , student or trainee might like to consult 'experts' (as a 'tutor') in his or her study, and use CAI programs which can simulate an expert. Such CAI software, containing an expert system is called an 'intelligent CAI' (ICAI). However, expert systems in CAI or CAL context have not yet been realized, but this will undoubtedly be developed in the near future. For the discussion it is important to discern at least three kinds of expert systems:
1. 'Logical decision tree' systems, based on decision rules (if-then-else-like
rules) which are similar to methods for determination of a plant with the
help of a flora.
2. Systems which use calculation of probability with the help of the
method by Bayes by which, for example, the probability of a symptom
of an illness is turned into a probability of that syndrome when such a
symptom is found.
3. Systems which make use of special languages like Prolog, Smalltalk
and LISP.
In chapter 1 the basic idea is described which underlying the second method described by a.o. Lusted (1968) and De Dombal (1972), using the method by Bayes. This is done on the basis of an example from the practice of a medical specialist with four syndromes in relation to seven symptoms. A non-medical subject could also have been chosen, but a medical example is perhaps more appealing, because it is a question of a diagnosis in the most ordinary sense of the word. With this expert system a diagnosis can be given (automatized) when a student answers the following questions:
1. Are you short-winded when straining? yes
2. Do your legs hurt? no
3. Do you have pain when straining? yes
4. Are your feet swollen at night? yes
5. Is your body blue? no
6. Does it hurt in the middle of your breastbone? yes
7. Do you feel heavy pressure on your chest? no
The computer program PHYSICIAN / ARTS answers:
The most probable diagnosis is Asthma Bronchial with a probability of 0.575.
The computer program containing the expert system has been developed by the Universiteit Twente, and is called PHYSICIAN / ARTS. There are three different versions. The first version was realized on an Apple II microcomputer and is used in a lab for the third year subject 'CAL-technics' for students of the University of Twente. The second version was developed on a Macintosh computer of Apple in Macintosh Pascal and it only uses texts (see figure 14.1). The third version was also developed on Macintosh and it has the special feature that the calculated chances are always graphically represented after answering a question. This provides good visual support during diagnosing. This option is also important for 'gauging' the expert system. It is not yet quite clear how expert systems will acquire a place in education. The possibilities for the training of medical students and laboratory workers are enormous. For non-medical education there are also plenty of possibilities. Think of a training where diagnosing has to be learned, like analyzing break downs in cars or computers: so in technical education; in determination as with a flora or fauna in microbiology: at laboratory colleges; in classification and making automatic divisions: in financial, economic or judicial fields; in pattern recognition; in geology, meteorology, etc.
At the University of Twente not only a medical example has been developed of an expert system based on this method, but also one from biochemistry that can differentiate between acute hepatitis, jaundice caused by a blockade in the bile, cirrhosis and liver metastasis, and an example from sociology that can differentiate between various political convictions.
The computer simulation program HISTORY TAKING / ANAMNESE is especially designed by the University of Utrecht to give students the opportunity to train themselves in taking the anamnesis (the interview between doctor and patient) of any given patient with a certain clinical picture. The program has five simulation patient cases. Before the student takes an anamnesis of a real patient, the student is given a chance to learn to survey the anamnesis for any given clinical picture. With this program it is possible to simulate in an easy (and cheap) way a doctor-patient environment. If a certain computer simulation patient is not relevant or has been dealt with, then another patient can be chosen immediately. The student is able to use the directly accessible patient data bank of the anamnesis training program, which is simpler than with an ordinary simulation patient. Beside a student can use this program the moment he or she is ready for it.
The program HISTORY TAKING / ANAMNESE introduces the student to the
classical construction of the anamnesis, in which it is precluded that
questions are asked in accordance with a choice hypothesis.
Other ways of taking an anamnesis are particularly not excluded while
the program is used. In the program the emphasis is not on problem solving
but on getting a complete survey of the anamnesis. The program is divided
into and based on principles according to the textbook of 'Anamnese en
der fysische diagnostiek' (Anamnesis, history taking and physical diagnostics)
by Formijne and Mandema (1976). According to this method it is important
that the student has to go through all the tracts ('chapters') of the anamnesis
before he or she formulates (a) provisional diagnosis(diagnoses). The student
has to pay attention to the complaint with which the patient has come.
Subsequently the patient is questioned extensively (from top to toe).
With this program there is a system of messages that offers instruction
at certain unexpected moments, accoording to Boolean logic. The presentation
of this kind of messages depends on the question, the patient and
his or her answers. The program also has a system of operations for self-testing
which are shown to the student at seemingly random moments and which have
to be answered before proceeding. The results of it come back at the end
of a program series in the evaluation of a session.
The computer simulation program HISTORY TAKING / ANAMNESE has originally
been developed in Utrecht, at the Rijksuniversiteit Utrecht (Min, 1978;
Min and Epraïm, 1979) at the Instituut voor Nucleaire Geneeskunde
(Institute for Nuclear Medicine) in cooperation with the Kliniek voor Inwendige
Geneeskunde (Clinic for Internal Medicine) of the Medische Faculteit (Faculty
of Medicine), Academisch Ziekenhuis Utrecht (AZU) (University Hospital).
The project 'Development of a program for self-tutoring for differential
research', from which this anamneses training program came, served as an
inventory of the possibilities which the computer has to offer in differential
diagnostics (Min, 1982).
The computer simulation program MacCOAT is especially designed by the University of Utrecht in cooparation with the University of Twente, to give students the opportunity to train themselves in taking the anamnesis just as with the program HISTORY TAKING / ANAMNESE (see figure 14.2).
With this program it is possible to simulate a educational doctor-patient environment on the same way as above. The student is able to use the directly accessible patient data bank of the anamnesis training program. The program has 4 simulation patient cases. With this program there is also a system of messages that offers instruction at certain unexpected moments, accoording to Boolean logic. The program is writen in LPA MacProlog.
The computer simulation program PLC for the subject mechanical technics has been developed within a team project between the Stichting voor de Leerplanontwikkeling (SLO) (a foundation for the development of curricula) and the Universiteit Twente. This collaboration wants the University of Twente to develop a number of computer simulation programs together with SLO on computers furnished by SLO. These programs will subsequently find their way into education via SLO.
This simulation program has been developed to introduce students of the LTO (a lower form of technical education) to the operation process in computer-operated apparatus. For it appears that students who have left the LTO are often confronted with such apparatus in their jobs, and are not or insufficiently prepared for it.
Introduction
At junior and senior secondary technical schools learning how to handle
and program so-called PLC. A PLC is an important aim. PLC stands for 'programmable
logical controller', a programmable apparatus that can operate various
other devices. These PLC's are present in industry in various types and
shapes. It can be regarded as a (micro)computer that can be programmed
on the spot to operate various machines. The 'programming language'
is characterized by a high degree of visualization and the ability to link
objects in a certain way. The object can be supplied with functions, an
activity and a condition for an activity.
At most schools there is only one complete PLC. By a complete PLC is meant: a collection of mechanical pneumatic parts such as pistons, relays etc., which can be mounted in various positions on a large mounting board causing all kinds of activities, provided they are operated by a PLC. This is called a PLC pneumatics set. The SLO at Enschede and the Universiteit Twente participated in this project. Wolters (1986) designed and realized the first version of the computer simulation program PLC on Macintosh. The second version with animation was realized by Dekker, Van Doorn and Pikaart with CAILIB, version 2.0 (see figure 14.3). Moen (1987) has meanwhile realized an MS-DOS version. This was programmed by an external company under direction of the SLO according to Wolter's design. (Wolters, 1987; Min, 1987; Dekker, Van Doorn and Pikaart , 1989)
Educational value
Usually schools can only afford to buy one completely equipped PLC.
Consequently pupils can only practice once or twice a year with the equipment.
Programming a PLC requires a great deal of practice in conceptual thinking,
and the ability to develop new programs independently and tracing the errors
in the programs that have been made. Most schools in the Netherlands use
the PLC made by the french firm Telematique. The equipment consists of
a PLC and a board that can be fixed on the wall on which cylinders and
pistons have been mounted that can be interconnected and influence each
other. All this is controlled by the PLC. In order to make the most of
the few times the PLC is used, a simulation program called PLC has been
developed and realized to train the pupils. With this computer simulation
program pupils can be optimally prepared for work in a practical training
room with the real apparatus. Beside the fact that many more pupils are
now busy with the problematic nature of a PLC and its programming, the
quality of the exercises has also increased. The pupils can practice how
to make programs independently. As was said earlier, the pupil is actually
'modelling CAL' with this simulation program.
For designing and filling in a function diagram (a program) involves
trying to realize a control model based on a conceptual model. The control
program ('the model') can be programmed on two levels. Level 1 is programming
the function diagram and level 2 is the filling in of the ladder diagram
with commands and conditions. These can be indicated by clicking in graphical
and numerical symbols. Thus the computer simulation program PLC has become
a sort of editor.
To computer simulation program PLC teaches the students how to handle a system with maximally three pistons and 10 actions (steps) and a function diagram and a ladder diagram in order to control the chosen installation with the relevant pistons. Two cases have been developed with the PLC program, one in order to learn how to program a stamping installation with two pistons and another for a 'lifting device' with three pistons. Dekker van Doorn en Pikaart (1989) realized a version of PLC (PLC animation) including a complete animation of a lifting apparatus. If the student programmed the lifting device well, this is represented on the screen in the form of an animation. This forms the end of the lesson with the computer simulation program PLC and is actually 'overall feedback' for the student.
Results
With the computer simulation program PLC the following goals can be
reached:
For a survey of the computer simulation program PLC see figure 1.7 in chapter 1
The computer simulation program PLC has yielded good results. It has been tested in three forms of a junior and senior secondary technical school. The accompanying paper materials were approved of. However, SLO has not yet published the three different versions of the PLC program. But research at the University of Twente to achieve an even better concept and userinterface continues.
Introduction
At the Waterloopkundig Laboratorium (Hydrodynamic Laboratory) in the
Flevopolder, Marknesse in the Netherlands, a mathematical model is used
that can calculate the course of the flow and the sediment in a
river or canal. For example it can calculate the level of the water
lines from the movements of the tides at the river mouth. It can also calculate the influence of rain in the hinterland. The mathematical model is called SEFLOW. The training program has been made by the University of Twente
together with the Hydrodynamic Laboratory which has for this occasion lent
the model developed by them in order to have a complete computer simulation
program built around it that can be used for educational purposes. The
computer simulation program will contain a series of cases ascending in
measure of difficulty so that course members can be trained for the more
difficult phase later on in their study: the modelling process of any river
or canal itself. The first prototype has been delivered by the University
of Twente on an ATARI 1040ST with the GEM operating system.
Model
At the Waterloopkundig Laboratorium water flows are studied in rivers
and in the sea with a lot of physical, geometrical and mathematical models.
In practice the construction, calibration and validation of mathematical
models requires a lot of time and sometimes the results are not accurate
enough. The experience of the user plays an important role,
particularly with respect to his or her knowledge of physics. For this
purpose the program FLOWSIM has been developed (see figure 14.4). FLOWSIM
is an interactive training program to gain insight into the response of
flow through networks, such as rivers and canals. Up till now it has been
constructed as a 'shell' around the FLOW part of SEFLOW and in future it
can be extended to any program.
Results
FLOWSIM enables the user to exercise real simulations on various models
while not being confronted with difficult and time-consuming aspects such
as model schematization, data-input, preprocessing and postprocessing.
By introducing a change into the system and running a new computation,
the influence on water levels and flow velocities will be illustrated.
In this way the user gets a feeling of the system's response to changes
in boundary conditions, bottom roughness, inclination of the waterway,
and so on. An instruction manual, providing the elementary physical background,
will guide the user during the exercises. The program is now (1991) available
on an MS-DOS computer with an GEM operating system. (Zwart, 1989)
Introduction
At Fokker Space and Systems, in the Netherlands (Amsterdam) a
mathematical model is used that can calculate the position of a satallite
in space. It is used for the dutch satallite that will be ready in 1993/94.
Ter Hedde developed an animation version for training
purposes and describes a prototype, named INSAN / SATSIM on a VAX computer
(see figure 14.5). (Ter Hedde, 1989)
Introduction
The computer simulation program HOSPITAL / ZIEKENHUIS was developed
by Botter, Universiteit Twente (1990) together with Talman and Houben,
Universiteit Limburg, Schoenmaker of the Educational Computer Consortium
(ECC) at Enschede.
The Universiteit Limburg provided a complete model that was working
well of bed occupancy in a hospital, written in a higher programming language.
Botter implemented this model into a SmallTalk learning and simulation
environment.
Model
The model of the computer simulation program HOSPITAL / ZIEKENHUIS
contains a system of bed occupancy. The model simulates part of a flow
of patients who are in a hospital with a capacity for 600 beds. The flow
of patients results in a specific bed, xxx occupancy depending on the control model (way of reservation).
This is what the trainee wants to find out. The model consists of the following parts:
The hospital model has five departments, with a number of beds in each
ward divided between rooms for two and for four persons. Emergency patients
are immediately admitted
to beds that have been reserved for this purpose in the ward, whereas
ordinary patients are put on a waiting list. Every day a number of patients
is released and a number of patients from the waiting list are admitted.
Here the principle holds true: First come, first served. The model registers
how many patients are admitted , refused and released, what the bed occupancy
is and how long one has to wait. Non emergency patients are also classified,
viz. patients who should treated before a certain date and those that are
not urgent, who can be admitted any time. Emergency patients are generated
through a Poisson division, with an average determined by the user. Non-emergency
patients are also simulated by means of a Poisson division and 25% of these
gets an urgency of 11 days at the maximum. Of every patient the age, sex,
ward (in fact determining the disease category) and the duration of the
stay in hospital is determined.
Program
The computer simulation program HOSPITAL / ZIEKENHUIS has five wards:
Every ward has a number of beds that can be changed, divided between rooms for two and for four persons. There are an equal number of either. Every room has patients belonging to the same sex and the difference in age is 15 years at the most.
When a patient is admitted he or she is entitled to the means of the
hospital. The patient will occupy a bed for a certain period of time, and
will affect the pressure of work on the medical and nursing staff. In order
to control the effect of these data, a control model and a measure for
its efficiency has been determined. The measure for the efficiency
is determined by in how far the hospital's aim is achieved. The aim is
optimalization of the bed occupancy on the following conditions:
Results
The program HOSPITAL / ZIEKENHUIS was successfully programmed in Smalltalk. The results achieved during the design and realization phase are in short:
- Interaction with the hospital model is difficult to optimalize in Smalltalk.
It proved impossible to produce an equally perfect userinterface within
the framework of this project as with Mac THESIS software;
- For the time being procedure libraries do not allow the linking up
of all kinds of animation tools to the model, such as 'Toolbox' of
Macintosh;
- Smalltalk is too difficult a language to master reasonably within
six months. Once one knows the language it is relatively simple to make
a program based on earlier prototypes.
As a prototype the program HOSPITAL / ZIEKENHUIS can certainly be called a success. This results in further research into object oriented languages at the Universiteit Twente.
Note
The paragraph about the computer simulation program SEFLOW / FLOWSIM,
paragraph 14.5, is original writen by W.J. Zwart & F.B.M. Min.
The paragraph about the computer simulation program SATSIM, paragraph 14.6, is original writen by R. ter Hedde & F.B.M. Min.
The paragraph about the computer simulation program HOSPITAL / ZIEKENHUIS,
paragraph 14.7, is original writen by B. Botter & F.B.M. Min.