To Main Proceedings Document |
Teaching System Dynamics and
Systems Thinking in Austria and G ermany
Dr. Giinther Ossimitz
University of Klagenfurt, Austria
Department of Didactics of Mathematics
A-9020 Klagenfurt, Universitatsstr. 65; Austria/EU
+43-463-2700-427; Fax +43-463-2700-427
http://go just.to/go
ossimitz@ bigfoot.com
Abstract
This paper discusses the emergence of system dynamics/systems thinking (SD/ST) teaching in
different countries. A special focus is placed on efforts made to introduce the MODUS
software in Germany and on the introduction of systems-thinking ideas in Austria's
mathematics curriculum in the early 1990s. In chapter 4 the fascinating relation between
systems thinking and system dynamics is discussed in some detail, followed by my own
definition of systems thinking. In the final chapters the main results of four empirical studies
concerning the development of systems thinking skills through teaching system dynamics are
summarized.
1 The Emergence of System Dynamics/Systems Thinking Teaching around 1990
In the late 1980s and early 1990s a number of initiatives to establish system dynamics or
systems thinking teaching (SD/ST teaching) emerged independently in different countries
worldwide.
¢ In the United States these efforts were focused around the revolutionary STELLA
software for Apple Macintosh computers. In several school districts teaching SD
modeling with STELLA was established, aiming for the development of systems thinking
skills. Among the first were the Brattleboro Union High School in Vermont or the
Catalina Foothill School District in Tucson, Arizona. Projects like STACI (Systems
Thinking and Curriculum Innovation, Mandinach 1989) or CC-STADUS (Cross-
Curricular Systems Thinking and Dynamics using STELLA, Waters Foundation 1996)
indicate the strong intention to promote systems thinking skills by using SD modeling.
* In the Netherlands Piet van Blokland developed the equation-oriented simulation system
VUpbyNamo!. Together with Douwe Kok, van Blokland introduced VUDYNAMo-based
system dynamics modeling at several Dutch schools in the late 1980s (Blokland/Kok
1989).
¢ In 1990 a German version of VUDYNAMO was introduced in the Hamburg school district
by Friedhelm Schumacher (Blokland/Schumacher 1990a, 1990b).
¢ In 1988 the Landesinstitut fir Schule und Weiterbildung (LSW) at Soest (the regional
school authority for Nordrhein-Westfalen - Germany) undertook a serious effort to
develop and promote Mopus, a graphics oriented software tool for systems modeling and
simulation for secondary schools. Within the MopDUS project teaching materials for
different subjects have been developed. Moreover extensive empirical research
conceming teaching system dynamics with Mopus and the development of systems
thinking abilities have been undertaken by Klieme and Maichle (Klieme/Maichle 1991,
1994).
¢ In Austria an initiative by Gerhart Bruckmann, a member of the Club of Rome,
(Bruckmann 1987) led to the implementation of a special section Untersuchung
vernetzter Systeme ("Investigation of interrelated systems") in the national mathematics
curriculum at 11th grade for the Realgymnasium (a special type of high school) in 1991.
All these initiatives (with the exception of the last one) were focused around a specific
system dynamics software product. They were typically of the bottom-up style, trying to
establish a small nucleus of SD/ST teaching first and extending this later. In most cases
enthusiastic individuals or small groups were behind these initiatives.
The only exception was Austria: here SD/ST teaching was implemented as a curri-culum
section first, hoping that this would trigger a subsequent teaching. Moreover the Austrian
initiative was the only one which did not rely on a single SD software product.
2 The MODUS Modeling Software in Germany
The modeling software MoDUS was specially designed for teaching systems oriented
modeling and simulation at secondary schools. It was developed under the name MOBILE at
the German Institute for Distant Teaching and Research (DIFF) at Tubingen by a group
leaded by Wemer Walser and Joachim Wedekind. (Walser/Wedekind 1991, Wede-
kind/Walser 1992). MopDUs was designed for PC computers and introduced a non-standard
graphical user interface with mouse and windows technique, emulated in a DOS
environment.
The graphical syntax of Mopbus was different from the stock-flow-diagram technique
invented by Jay Forrester (1961). The MobDuS-syntax did not distinguish between inflows
and outflows: both were considered as effects upon the stock variable, differing just in the
sign. This resulted in an unfamiliar way of representing elementary systems: in MODUS
stocks could only be changed by
"change variables" (like z
change_pop in Fig. 1). Both MODUS:style:
inflows and outflows have to be Ts births
Tepresented as change variables <_<
just with different signs, but
always with a logical arrow population change_pop
towards the stock variable. deaths
Actually the Mopus style of stock-flow-diagram style (Forrester):
diagramming proved to be
counterintuitive and much harder a =e
to grasp for students than the oe =
Forrester-style stock-flow population
diagrams. When MODUS was births deaths
commercially released in 1992 5 . .
after about four years of Fig. 1 Mopus vs. stock-flow-diagramming style
development and empirical testing, its DOS-based non-standard graphical user interface was
de facto already outdated by the MS-Windows revolution. In those days the first versions of
POWERSIM with a standard Windows-interface made it clear that Mopus would never
achieve practical status as a teaching tool.
Nevertheless the efforts around MODUS were important. A significant amount of
development of teaching materials and empirical research accompanied the Mopus project.
The empirical research was done in two big studies by Eckhard Klieme and Ulla Maichle
from the Institute for Educational Research (Institut fiir Bildungsforschug) at Bonn.
In the pilot-study Klieme/Maichle (1991) 180 students in 8 classes at grade 9 and 10 were
studied to see the implications of a teaching sequence of about 15 - 20 hours introducing
Mopus. The teaching was done using different materials that had been developed at the LSW
Soest for mathematics, biology, chemistry and social science classes.
The testing of the students was done in a pre-test - post-test design using written tests which
were not known to the teachers. The test covered several themes with a clear focus on
modeling and systems thinking abilities. The tasks to be done at the pre-test and post-test
corresponded to a very high degree (with the exception of some Mopus-related questions,
which were not asked at the pre-test.) One main achievement of Klieme and Maichle was the
development of reliable testing tasks that would allow a fair measurement of systems
The Hilu-tribe
The African Hilu tribe breeds cattle. The income of the Hilus depends upon the number of animals
they can sell per year. The bigger the herd, the more animals are sold and the higher the annual
income of the Hilus becomes. The more money they earn, the more they can invest in their recently
built bush hospital for medicines and instruments.
Since rainfalls are rare, the Hilus have drilled a deep-water-well and had installed a water irrigation
system. Increasing watering raises the moisture of the grasslands. This has pros and cons: More
moisture lets the grass grow better and the cattle can grow, too. On the other hand the moisture
supports the propagation of the dangerous tse-tse-fly. This fly spreads the dangerous cattle
solemnia disease, which every year infects a part of the herd. With an increasing number of tse-tse
fliese more cattle die on this disease. If the irrigation is reduced, both the food supply of the cattle
and the propagation of the tse-tse-fly are reduced.
Try to sketch the interrelations described here in a diagram in such a way that one can see the most
important aspects at a glance!
Fig. 2 The Hilu scenario in Klieme/M aichle (1994) and Ossimitz (1994)
thinking abilities both in the pre-test and post-test. Fig. 2 shows an example.
The Klieme/Maichle studies greatly affected my own empirical research on the development
of systems thinking abilities (Ossimitz 1994, 1996, 2000). I adopted most of the tasks for
measuring systems thinking abilities that have been developed by Klieme and Maichle for my
own investigations.
3 Austria: An Exceptional Curriculum Design
The emergence of system dynamics teaching in Austria around 1990 was quite different from
all other initiatives of that time that I know. In Austria teaching system dynamics was not
introduced by initiatives at the "teaching front", but in a top-down manner during a revision
of the mathematics curriculum for the 9th-12th grade of the Realgymnasium’. At grade 11 a
section Untersuchung vernetzter Systeme (Investigation of interrelated systems) was
introduced. This section states explicitly that in the mathematics classes at grade 11 students
should gain systems thinking abilities by analyzing systems from different fields like
economy, ecology, biology or physics. The curriculum stresses the importance of different
modes of denoting or diagramming systems, using causal loop diagrams, stock-flow diagrams
or formal equation-style notations. Although intended, the system dynamics method is not
mentioned explicitly, so that alternate modeling styles like using a spreadsheet could be used.
The curriculum simply emphasizes that the investigations should lead "finally" to a numerical
evaluation and simulation, preferably using a computer.
An important intention of the curriculum section is that the underlying assumptions, and
limitations of the modeling process and the resulting scenarios should be reflected and
discussed - something that is rather uncommon in mathematics classes.
The curriculum section Investigation of interrelated systems was not restricted to any specific
modeling and simulation technique. This was done to give the teacher some freedom and to
help to keep the curriculum section fit for later technological innovations.
In order to make the intentions of the curriculum section more explicit and to support the
authors of the mathematics textbooks and the Austrian mathematics teachers with prototype
models and practical SD teaching knowledge, the book Materialien zur Systemdynamik
("Materials for Teaching System Dynamics" - Ossimitz 1990) was published. This book
supports both the "hard" quantitative and the "soft" qualitative modeling paradigm. In the
field of quantitative modeling and
simulation the system dynamics approach | 3 = 6 6?P al
is clearly preferred. Many of the simple (P2) (Pa)
discrete models that do not require Runge-
Kutta integration are explained both in the (2) (a)
VUDyYNAMO syntax and as spreadsheet
models, so that teachers who are familiar Fig. 3 stock-flow diagram from an Austrian
with spreadsheets do not necessarily need mathematics schoolbook
a system dynamics software.
A disappointing aspect of the Austrian experiment to establish systems thinking in math
classes via curriculum innovation was the way in which the new curriculum section
Investigation of interrelated systems was interpreted by the authors of the three major appro-
ved mathematics text books for the 11" grade. Only one of the textbooks made any serious
attempt to introduce stock-and-flow diagrams. Fig. 3 shows the result: all variables bear
typical mathematical one or two-letter
variable names. This reductionism makes Population
it hard to use the models for discussing S
the design of a particular stock-flow- ~————
diagram. growth 4-) Capacity
Moreover the very same stock-flow- rowth ne
diagram in Fig. 3 was used in that actor fraqtion
textbook for modeling logistic(!) growth. space
By setting PA := P2 + a instead of _.
PA :=P-a Fig. 3 represents a Verhulst- Fig. 4 A reasonable logistic arowth model
like logistic population model of the type dP/dt = zP - aP? (see Richardson 1991, p. 32). Of
course no serious SD modeler would take a structure like Fig. 3 for modeling logistic growth.
He or she would argue that such a model should contain explicitly important aspects of the
teal system, such as its capacity or the fraction of unused capacity. An acceptable stock-flow-
diagram might look as the one in Fig. 4, using equations like
growth := growth factor - fraction of free space - Population and
fraction of free space := 1- population/capacity
The other two mathematics textbooks treated the basic ideas of system dynamics and systems
thinking even worse: one of them displayed just a single stock-flow diagram of an elementary
capital - interest growth without any explanations. The only commentary was an exercise
"Try to find out what this diagram could mean!" The remaining 25 pages of the chapter were
focused exclusively on some mathematical theory of difference equations. In the third
textbook just a single stock-flow-diagram of population growth like in Fig. 2 is discussed in
some detail.
In 1993, two years after the official introduction of the curriculum section Investigation of
interrelated systems, Franz Schloglhofer investigated in his doctoral thesis (Schléglhofer
1993) the extent to which this section has already penetrated mathematics teaching in Austria.
The results were mediocre.
The experience made in introducing system dynamics and systems thinking in Austria at the
level of teacher-oriented initiatives was by far more encouraging. In-service teacher courses
of about one week were held in 1993 and 1994 and had a considerable effect upon the
teachers’ thinking. Additional empirical research (Ossimitz 1994, 1996) showed that the
induced teaching actually had some impact upon certain dimensions of the students' systems
thinking abilities.
4 The Relation between System Dynamics and Systems Thinking
In order to clarify this last argument it would be useful to give a short overview about the
fascinating relation between system dynamics and systems thinking. In the early writings of
Jay Forrester no indication of the term "systems thinking" can be found. The very first
occurrence of a connection between system dynamics and systems thinking that I could find
was is in a paper by Ellen Mandinach (1989) about the STACIN project. The mere acronym
Systems Thinking and Curriculum Innovation Network indicates that the notion of systems
thinking plays a core role in this project. Mandinach reports on four courses using STELLA
software at Brattleboro Union High school in Vermont. Mandinach (1989, p 222) writes:
"Systems thinking is a scientific analysis technique given prominence by Jay Forrester
and his colleagues at the Massachusetts Institute of Technology. Work on computer
modeling of systems thinking started well over 30 years ago with early models
focusing on urban growth and development and global patterns of the consumption of
natural resources."
A few pages later Mandinach concludes:
"As defined here the systems thinking approach consists of three individual but
interdependent components: system dynamics, STELLA and the Macintosh."
(Mandinach 1989, p 225)
It is rather obvious that Mandinach presents an idea of systems thinking which equates
systems thinking and system dynamics to a high degree. Actually Jay Forrester himself never
had such an intention. On the contrary: in one of his later papers System Dynamics, Systems
Thinking, and Soft OR (Forrester 1994) the founder of system dynamics addresses this issue
explicitly. First he gives six "System dynamic steps from problem symptoms to
improvement", which range from "Step 1: Describe the System" to "Step 6: Implement
changes in policies and structure". Then Forrester identifies several procedures which might
be helpful for doing Step 1. One of these procedures (among others) is systems thinking. He
writes rather critically:
" “Systems thinking” has no clear definition or usage. ... Some use systems thinking
to mean the same as system dynamics. ... “Systems thinking” is coming to mean little
more than thinking about systems, talking about systems, and acknowledging that
systems are important. In other words, systems thinking implies a rather general and
superficial awareness of systems. Systems thinking is in danger of becoming one
more of those management fads that come and go. The term is being adopted by
consultants in the organization and motivation fields who have no background in a
rigorous systems discipline." (Forrester 1994, pp 10-11)
Forrester accepts systems thinking as a kind of "door opener" for rigorous system dynamics
modeling; but he refuses the identification of system dynamics with systems thinking. He
fears that the formula SD = ST might allow non SD-oriented "systems thinkers" to enter the
field and damage the serious reputation of SD modelers.
So what made Ellen Mandinach believe that Jay Forrester is the father of systems thinking? I
think that at least a part of the answer can be found in Barry Richmond, the ingenious
inventor of the revolutionary STELLA software and the main head behind the great STELLA
Academic User Guide (Richmond/Peterson/Vescusco 1987). Richmond propagated SD
modeling with STELLA from its earliest days quite successfully under the label of "systems
thinking". This might have triggered Mandinach's view of Forrester as the father of system
dynamics = systems thinking.
In the papers Systems Thinking: Four Key Questions (Richmond 1991) and Systems
Thinking: Critical Thinking Skills for the 1990's and Beyond (Richmond 1993) Richmond
addresses the relation between system dynamics and systems thinking extensively. The four
key questions in Richmond (1991) are:
1. What is Systems Thinking?
2. Why is Systems Thinking needed?
3. What works against the adoption of Systems Thinking?
4. What can be done to facilitate the adoption of Systems Thinking?
In the introduction Richmond writes:
"Systems Thinking, A Systems Approach, System Dynamics, Systems Theory, and
just plain 'ol "Systems" are but a few of the many names commonly attached to a field
of endeavor... As I prefer the term "Systems Thinking", I'll use it as the single
descriptor for this field of endeavor." (Richmond 1991, S. 2).
This reads as if Richmond is not very careful to distinguish between system dynamics and
systems thinking and that "systems thinking" seems to him as a nice term to coin as a trendy
description for system dynamics modeling.
In Richmond (1993) he gives a more explicit definition of systems thinking. Richmond
discusses seven basic "systems thinking skills". A closer look at these skills clearly reveals
that these are just the skills which are useful for doing successful system dynamics modeling.
Of course most of them have some relevance outside SD modeling, too, but taken together
these skills are pretty well what I would say that a good SD modeler would need. For some of
the skills, like "generic thinking", "operational thinking" or "continuum thinking", it is even
somehow hard to see how they would make sense outside an SD modeling context (For
details see Ossimitz 2000).
The essence of Richmond's view about the relation between SD and ST can be found in his
paper Systems Thinking - Let's Just Get On With It (Richmond 1994), which might well be a
reply to Forrester (1994). In this paper Richmond cites Forrester's belief that systems thinking
is only about 5% of the system dynamics modeling process. "The other 95 percent lies in the
tigorous System Dynamics-driven structuring of models and in the simulations based on
these models" says Forrester, cited by Richmond. Metaphorically speaking, in Forrester’s
opinion systems thinking is just a small territory on the system dynamics globe. Richmond
gives his own picture by drawing a system dynamics globe with an ‘systems thinking
atmosphere’ around it and writes: "Systems Thinking is ‘System Dynamics with an aura’."
(Richmond 1994, p 4)
Richmond's way of defining systems thinking has several implications. First it makes systems
thinking achievements rather easy: gathering any system dynamics modeling knowledge
means per definitionem the acquisition of systems thinking skills. The second implication of
Richmond's definition is that empirical measuring of systems thinking skills can be boiled
down to the measurement of system dynamics modeling skills. Under this assumption it is
just a little bit weird to put such a kind of measurement in a pre-test - post-test design with
two identical tests: before being taught about SD modeling it can be expected that any student
will achieve very poorly in any SD modeling achievement task. At first sight this looks rather
trivial: before being taught about SD modeling the students have no idea about SD modeling
and systems thinking - so what? On closer inspection the close identification between SD
modeling and systems thinking implies that without learning some SD modeling technique
there are no (measurable) systems thinking skills at all! We can put it more dramatically:
identifying systems thinking and system dynamics as closely as it is done by Richmond
would imply that before 1958, when Jay Forrester invented the SD modeling method, there
were no systems thinkers at all in the whole world! I am sure that such an implication would
be considered a little bit too heavy, even by Barry Richmond himself.
5 What is Systems Thinking?
Pondering the consequences of Richmond's ideas brings us to the point that there is a need for
a definition of systems thinking which is somehow independent of the system dynamics
modeling approach.
When looking in the literature for such a definition, it is surprising how little is to be found,
although the term systems thinking ("systemisches Denken", "vernetztes Denken" etc.) is a
widely used phrase both in international and in German literature. Yet it is hard to find a
concise definition of "systems thinking". Let me give some examples of my findings. Klir
(1991, p 19) writes in his monumental book Facets of Systems Science:
“Systems movement emerged from three principal roots: mathematics, computer tech-
nology, and a host of ideas that are well captured by the general term systems thinking.”
Although this citation suggests that systems thinking is something very fundamental for Klir,
this single sentence is all that Klir says explicitly about systems thinking. Klir gives no closer
hint about what he means by “systems thinking”.
The German cognitive psychologist Dietrich Dorner (1989, p 308ff) says in his book The
Logic of Failure (Die Logik des Mi8lingens):
“T hope I could clarify the fact that we cannot grasp what is often generally called «sys-
tems thinking» as a simple entity, as an individual, distinguishable ability. It is a bundle
of abilities, and essentially it is the ability to use our normal, sound reasoning according
to the circumstances of the individual situation.” Dorner (1989, p 308ff, translated by G.0.)
Here Dorner essentially reduces systems thinking to the formula:
systems thinking = complex situation + a thinking mode adequate to the situation.
6 A Definition for Systems Thinking
I would like to define four essential dimensions of systems thinking:
1) thinking in models: explicitly comprehended modeling
2) closed loop thinking: thinking in interrelated, systemic structures
3) dynamic thinking: thinking in dynamic processes (e.g. delays, oscillations)
4) steering systems: the ability for practical system management
6.1 Thinking in Models
From the viewpoint of Radical Constructivism (cf. eg. Glasersfeld 1995) thinking in models
is inevitable. Constructivism says that we can only think according to our pictures and views
of the world, which are necessarily models of the world itself. Now my point is that systems
thinking requires the consciousness of the fact that we deal with models of our reality and not
with the reality itself.
Thinking in models also comprises the ability of model-building. Models have to be
constructed, validated and developed further. The possibilities of model-building and model
analysis depend to a large degree on the tools available for describing the models. Choosing
an appropriate form of representation (e.g. causal loop diagram, stock-flow diagram,
equations) is a crucial point of systems thinking. The invention of powerful, flexible and yet
standardized descriptive tools was one of the main achievements of Jay Forrester. For school
purposes the representation forms of the System Dynamics approach have proven to be
successful. The causal loop diagram allows qualitative modeling, the stock-and-flow diagram
already gives key hints about the structure of the quantitative simulation model.
6.2 Closed Loop Thinking
People of the western hemisphere are usually very good in causal reasoning. If-then relations
are basic building blocks of our mind and our understanding of things. A foundation of this
kind of thinking is a strict distinction between cause and effect. In order to explain a
phenomenon we have to find its (probably single) “cause”. It is supposed that this cause does
exist and that the effect always can be observed whenever the cause is valid. Words and
phrases like “because”, “therefore”, “if - then” denote such thinking concepts in everyday
language. The mathematical analogon is the function-concept with one independent variable
(="cause”) and one dependent variable (="effect”). Accordingly the thinking in simple cause-
effect relationships might be called functional or linear thinking - in contrast to closed loop
thinking.
In interrelated systems we have not only direct, but also indirect effects. This may lead to
feedback loops. Feedback loops might be reinforcing (positive) or balancing (negative). The
arms race between the superpowers was an example of a reinforcing feedback loop.
Americans said: Because of the armament of the Soviets we have to build 1000 new
missiles”. The Soviets said: “We have to increase our strategic arms force, because the
Americans have built 1000 new missiles.” This increase in the Soviet Army Forces led to
further armament on the American side... and so on. Each side viewed the other side as the
cause. In a global perspective a distinction between cause and effect is no longer possible.
Once you have entered a vicious circle, you can no longer identify a single cause for the
whole process, since any effect also affects the cause. A proper understanding of feedback
loops requires a dynamic perspective, in order to see how things emerge over time.
Interrelated thinking is a kind of thinking which takes into account indirect effects, networks
of causes and effects, feedback loops and the development of such structures over time.
Interrelated thinking also requires adequate representations: the causal loop diagram is the
simplest and most versatile tool for denoting interrelated issues.
6.3 Dynamic Thinking
Systems have a certain behavior over time. Time delays and oscillations are typical features
of systems, which cannot be observed without the time dimension. Even the simple task of
keeping the temperature constant in a (simulated) cooling house is for many subjects a
difficult task, because changes in the temperature would require some time before they
became effective (see Dormer 1989, pp 200ff). Considering only the present state of the
temperature as a guideline for adjustment might lead to serious overreaction, which might
take even a rather inert system like a refrigerated warehouse out of control.
Dynamic thinking also means foreseeing (possible) future developments. A mere retro-
spective view of past developments is insufficient for the practical steering of systems - or
would you trust a car driver who makes exclusive use of the rear mirror in order to determine
where to steer the car? Often simulation models are helpful or even necessary in order to
foresee future developments - especially when reality emerges rather slowly.
6.4 Steering a System
This brings us to the fourth core aspect of systems thinking: the practical steering of systems.
Systems thinking also always has a pragmatic component: it deals not just with contem-
plating the system, it also is interested in system-oriented action.
One of the most fundamental and most important questions of practical systems management
is: Which of the systems components are subject to direct change? In a social system it is
often impossible to change the behavior of others directly, one can only change one’s own
behavior. In an economic system the producer usually has no direct control over the market.
Marketing activities are usually actions on the supply side in order to induce the desired
reaction on the demand side.
7 How can Systems Thinking Skills be Developed?
This is a very hard question. There are a number of different approaches (or claims), how “it”
could be done. Let me give an overview of some possible answers:
¢ Sensibilization for systems aspects by information campaigns for the general public (see
F, Vester’s exposition Unsere Welt - ein vernetztes System (our world - an interrelated
system, Vester 1986) or the works of D. & D. Meadows concerning The Limits to Growth
(Meadows 1972).
¢ Dorner (1989, p 307ff) suggests computer-simulation games (like Tanaland or
Lohhausen), in order to learn systemic thinking and action.
* Group-dynamics oriented approaches try to develop systemic skills as holistic
encounters in special seminars (e.g. the Tavistock concept for the development of
systemic management abilities).
¢ Some curricular concepts try to develop systems thinking skills via explicit teaching at
schools. Examples of this are the Austrian Schools Mathematics Curriculum section
Investigation of interrelated systems (Untersuchung vernetzter Systeme) or, on a more
comprehensive scale, curricular projects like CC-STADUS (Cross Curricular Systems
Thinking and Dynamics Using Stella or STACI (Systems Thinking and Curriculum
Innovation Network) in the USA.
For the rest of this paper I will confine my considerations to the curricular-oriented efforts. I
will try to summarize several empirical studies devoted to this subject.
8 Can Teaching System Dynamics trigger Systems Thinking A bilities?
This has been a core question of a number of empirical studies undertaken by Klieme/-
Maichle and by myself. I will give an overview about the design and the main results of the
following studies:
(1) Klieme/Maichle (1991): Erprobung eines Modellbildungssystems im Unterricht
(Evaluation of a Model Building System in the Classroom).
(2) Klieme/Maichle (1994): Modellbildung und Simulation im Unterricht der
Sekundarstufe I (Modeling and Simulation in Grades 9 and 10)
(3) Ossimitz (1994): Systemdynamiksoftware im Unterricht (System Dynamics Software in
the Classroom)
(4) Ossimitz (1996): Entwicklung vernetzten Denkens (Development of Systems Thinking)
The design of these studies is summarized in Table 1. All studies addressed the following
questions: Can systems thinking be taught in an ordinary school environment? To what extent
can the System Dynamics method facilitate the development of systems thinking and action
skills?
In each study the students were tested before and after a System Dynamics teaching module
of about 10 - 25 hours. The tests were in writing and not known to the teachers. The tasks of
the post-tests were closely similar to the corresponding pre-tests. In both studies of
Klieme/Maichle and in the Ossimitz (1994) study the graphical simulation software MopuS
was used. (Since in Germany and Austria almost all schools were equipped only with DOS-
PC’s at the beginning of the 90ies, STELLA or some other Windows-oriented software could
not be used).
Study
Students/
Classes Grades Teaching Subjects Software Research Method
K/M91 | 180/8 | 9; 10 |math, biology, chemistry, soc. stud. | Modus |pre- & post-test
K/M94 | 240/10 | 9; 10 jeconomy, biology, information sci. | Modus |pre- & post-test, selected videos
Oss 94) 7/2 9; 11 [mathematics Modus |pre- & mid- & post-test + interviews
Oss 96| 130/7 | 9to 12 |mathematics, inform. sci., physics | Powersim|pre- & post-test
Table 1: Design of the Klieme/Maichle (1991, 1994) and Ossimitz (1994, 1996) studies
The general results of all these studies were:
Most students and teachers considered the system dynamics teaching modules very
interesting. (In some classes it was the first attempt both for teachers and students of
computer-assisted teaching).
The way in which a systemic situation given as a text was sketched graphically changed
considerably between pre- and post-test. Most of the students who learned about causal
loop diagrams used them in the post-test; whereas in the pre-test most students used
pictorial images or verbal summaries to sketch the systemic situation. Students who did
not see causal loop diagrams, but structural or stock-flow diagrams, often used this type
in the post-test.
The ability to understand a systemic situation given as a text was very good even for the
youngest students (aged about 14-15) in the pre-test. Thus in the Ossimitz (1996) study a
rather complicated text was used for the pre- and post-test.
To teach systems thinking requires a great deal of engagement on the part of the teacher.
The students’ advances in systems thinking skills depended basically on the teachers' own
motivation and mental models about systems. The students of open-minded, systems
oriented teachers achieved significantly better improvements in all the SD-related
measures we applied.
The modeling style of MoDUs is much harder to understand than the style of POWERSIM.
POWERSIM was also considered to be far superior to non-graphical modeling options like
VUDyNAMo or spreadsheet modeling by all those teachers who have seen all these
variants.
Measurement of systems thinking skills and their improvement is very difficult. “Systems
thinking is no general ability. To understand the dynamics of Modus-models is something
different than predicting effects in verbally described models.” (Klieme/Maichle 1994, p
76)
Particularly important results of the Klieme/Maichle (1991) pilot study were:
The interest of the students depended upon their readiness to work on the computer. This
readiness differed greatly. (For the majority of the students it was the first contact with a
computer).
The Klieme/Maichle (1994 pp 73ff)
study yielded the following special
results:
The software product MopDus had a considerable impact upon the teaching and the way of
modeling. Some of the students’ problems in the modeling process seemed to depend
upon the specific way that systems
are modeled in MoDUS. The logic 2°00
of Mopus seemed to be counter- IST titan
intuitive and a hurdle for many
students to understand structural
diagrams.
The students of teachers using a
highly directive teaching style got
better results concerning the [in vs _
achievement of model building in im it @ =
the post-test.
The achievement of the students Fig. 5 Typical sketch of the Hilu - scenario (pre-test)
depends more upon their motivation and prior experience with computers.
Through the work with Mopus the ability of model building could be significantly
improved. The ability to think systemically within the MoDuUs-models could be improved
only marginally.
The readiness of the students for experimental work and cooperation was significantly
improved by the software-supported teaching compared with the "usual" non-computer-
oriented education.
In the pilot-study of Ossimitz (1994) only a few students of each class were tested before and
after a System Dynamics module of about 10 hours (grades 9 and 11), using a part of the
Klieme/Maichle (1994) test. After each test the students were interviewed about their
answers. The test + interview combination was more successful in the determination of
thinking processes than a design without interviews.
The Ossimitz (1994) study also showed that the
use of causal loop diagrams (CLD's)can be taught
to ninth-graders astonishingly fast. In one class
the "pre-test" accidentally took place after the first
hour of SD - teaching. In this lesson the students
had been shown just two very simple causal loop
diagrams. They spontaneously used causal loop
diagrams to describe Hilu-scenario (Fig. 2) in
their pre-test. An example is shown in Fig. 6.
Without prior teaching most students just used
pictorial diagrams to describe the Hilu-situation.
Fig. 5 gives a typical example.
Fig. 6 causal loop diagram of the
Hilu-scenario (after seeing 2 CLD's)
9 The Development of Systems Thinking - Study (Ossimitz 1996)
The project Development of Systems Thinking ("Entwicklung vernetzten Denkens") was
undertaken in cooperation with six volunteering teachers of Austrian (mostly business-
oriented) high schools. First the teachers got a one-week introductory course about the
principal ideas of System Dynamics, systems thinking and systemic modeling and simulation
using the simulation software PowERSIM. Then the teachers were free to design a teaching
unit of approximately 20 hours on system dynamics modeling and simulation. They were
only required to document the unit and to use POWERSIM in some way. For a detailed
documentation see Ossimitz (2000).
The students were tested with a written test (of about 45 minutes) before and after the
teaching (the test was not known to the teachers). The test design was similar to the study of
Klieme/Maichle (1994). The pre- and post-tests consisted of two main tasks with some
additional questions.
The first task in the pre-test was to depict a variant of the "Hilu'-scenario in a picture or
diagram; in the post-test a similar "Mori"-scenario was presented. On a formal level, the
complexity and inter-relatedness of both scenarios were (almost) identical. Both scenarios
had been given additional complexity (several closed loops) compared with the text in Fig. 2,
because the Hilu-text Fig. 2 turned out to be no real challenge for almost all students aged 14
- 16 in the pre-tests of the Klieme/Maichle (1994) and in the Ossimitz (1994) studies.
In addition, students were asked about indirect consequences of some actions (like "What
effect does using more grass-fertilizer have upon the abundance of the "tse-tse-fly?"). In the
evaluation of this task, mainly the type and the systemic complexity of the resulting diagram
and the quality of the additional answers were evaluated.
The second task, called arguments and counter-arguments ("Argumente und
Gegenargumente"), was taken without change from the Klieme/Maichle (1994) study.
Following a given example, students should write down chains of arguments (like more
tourists more hotels more traffic problems less attractivity of the resort) for traffic
problems in a small rural town (pre-test) and tourist problems in a sea-side holiday resort
(post-test).
For the evaluation of both tasks the number of items (Elements in the graphs) and relations
(arrows, logical if-then relations) being stated by the students were counted. These basic
measures were used to calculate an index of complexity and an index of inter-relatedness.
These indices were used as indicators to measure the skill of designing interrelated systems.
The measured items were also correlated with basic variables like gender, age, grade in
mathematics, computer experience. About 40% of the students owned a private computer;
about 10% of all students worked more than 6 hours per week on a computer.
9.1 Results of the Pre-Test
The students used a wide variety of diagram-types to depict the Hilu-scenario (see table 2).
There was no significant correlation between the type of diagram and the age, gender or
mathematics grade of the students. The results for the arguments-and-counterarguments task
were similar: I could not observe any correlation between the complexity-index or the
interrelatedness-index with age, gender or mathematics- grade.
9.2 Results of the Post-Test
The diagrams the students used for sketching the Mori-scenario at the post-test were in most
cases considerably different from the pictures they used in the pre-test. Generally pictorial
and verbal descriptions decreased and the number of causal loop diagrams increased
significantly. In some classes (Table 2, teacher T3 and teacher T6), almost all students used
causal loop diagrams. An extreme to the opposite side was the class of teacher T1: One third
of the students of teacher T1 used pictorial descriptions even at the post-test. Moreover they
were the only who drew pictorials in the post-test. However, the circumstances in the class of
teacher T1 were disastrous: in this class the "project" lasted for just 6 hours. Half of this time
was spent with very general discussions about medical health care, reasons for early deaths,
the belief in astrology and the like. Only about 1'4 hours were spent in the computer room.
The rest of the time the teacher discussed a drug addiction model using transition matrices
(no system dynamics), which led to the mathematical interesting issue of fixed points in
iterated vector equations.
Type of picture/diagram pre-test: post-test: Mori-scenario
used for Hilu / Mori-scenarios Hilu-sc. (teachers T1 ... T6)
all teachers|all teachers] T1 | 12 | 73 | 714 | T5 | T6
pictorial or verbal descriptions 22% 6% 33% 6%
function charts (mostly Cartesian style) 5% 3% 8% | 6%
chain diagrams (lin. seq. of arguments)} 15% 3% 11% | 8%
tree diagrams (mostly 2 branches) 1% 7% 11% | 16% 6% | 4%
causal diagrams without loops 29% 19% 28% | 33% | 5% | 12% | 21% | 11%
causal diagrams with loops 171% 61% 11% | 42% | 95% | 71% | 67% | 83%
Total 100% | 100% |100% |100% |100% |100% |100% |100%
Table 2 Types of diagrams used to denote the Hilu/Mori scenarios
For the first task (Hilu/Mori-scenarios) most indicators for systems thinking in the post-test
were significantly higher (on the 95%-level) than in the pre-test. E.g. the average index of
interrelatedness of the Mori-diagrams was at about 2.29+0.12 (double standard error of the
mean) compared with 1.47+0.14 in the pre-test. Table 3 shows that this average result over all
students differs significantly for both tests according to the teacher. The students of teacher 1
had the worst result at the pre-test (1.01) and almost no increase between pre- and post-test
(1.18). The students of teacher 3 were below the average in the pre-test (1.39), but they got
the best result (2.83) in the post-test.
Index of interrelatedness Tl T2 T3 T4 | T5 | T6 | Total avg.
Pre-Test: Hilu-scenario (avg.) 1.01 | 1.67 139 1.70 | 1.45 | 1.73 147
Post-Test: Mori-scenario (avg.) 118 | 2.00 2.83 2.15 | 2.11 | 2.50 2.29
Table 3: Average index of interrelatedness by teachers T1 - T6
For the arguments-counterarguments task several of the indices to measure the level of
systems orientation (complexity, interrelatedness, number of items and number of causal
arrows) were significantly higher in the post-test than in the pre-test. Again there was no
correlation with age, mathematics grade or computer experience, but a great influence of the
variable "teacher". I observed that the index of interrelatedness was slightly higher for the
boys than for the girls (for both pre- and post-test); but this might be just a side-effect of the
fact that the best teachers had a higher percentage of boys in their class.
Some other interesting results of the Ossimitz (1996) study were:
* Two other classes (at different schools) were also tested as control groups (without SD
teaching between pre- and post-tests). There was no significant difference in the students’
behavior between pre- and post-tests. A test-retest effect (that the students lear from the
pretest and thus achieve better results in the posttest) could not be observed.
¢ The teachers reported very positively on their teaching projects. They said that the
Windows-oriented software POWERSIM was (unlike MODUS) very easy to learn - despite
the fact that only an English version of POWERSIM was available and that POWERSIM has
many features far beyond the elementary needs of teaching.
¢ The teachers had few problems in giving homework and written tests for the system
dynamics module.
* Most students had no prior experience in project-oriented teaching, which caused some
insecurity at the beginning. In the final review of the teaching project most students gave
very positive comments about project-oriented teaching and the simulation software
POWERS. Here are two statements by students (from their written feedback about the
project):
"This kind of teaching was new for me, but somehow I appreciated it. I would not like all
teaching to be like this, but now and then it would be fine; especially so that one can
understand the connections between certain things."
"T liked the project. It was fun that we could work on our own. | like it better when we
can summarize the items ourselves than when the teacher talks for an hour or so."
10 Conclusions and Summary
The project Development of Systems Thinking showed that it is possible to construct
indicators for the development of systems thinking skills and to measure some progress
through a teaching unit of about 20 hours. The central result of our study is that the
variable “teacher” has by far the greatest influence upon explaining the differences
between the pre-test and post-test achievements. The impact of the teacher proved more
important by far than the variables "age", "gender", "computer experience" or "mathematics
grade". Even the style of teaching (conventional vs. project-oriented) did not affect the
outcome. In retrospect this might not be so surprising, but in the design-phase I definitely did
not expect the teacher’s impact to be so strong.
Let me conclude with a remark about "learner-centered-learning", as proposed by Forrester
(1992) for teaching and leaming System Dynamics. I do not think that leamner-centered-
learning lessens the importance of the teacher. In my opinion the teacher is the key factor for
introducing and maintaining leamer-centered-learning at school. We cannot expect that most
students automatically have the same motivation and excitement as researchers in their
laboratory; thus it is the teachers’ role to induce and to guide the motivation of his or her
students.
11 Literature and Web-Resources
Blokland, P. v., & Kok, D. (1989). Dynamische Simulatie . Amsterdam: Vrije Universiteit.
Blokland, P. v., & Schumacher, F. (1990a). Handbuch zur Programmierumgebung VU-
Dynamo . Hamburg: Behdorde fiir Schule, Jugend und Berufsbildung.
Blokland, P. v., & Schumacher, F. (1990b). Modellbildung und Simulation mit VU-Dynamo .
Hamburg: Amt fiir Schule.
Bruckmann, G. (1987). Zur Didaktik der Systemdynamik. In W. Dorfler, R. Fischer, & W.
Peschek (eds), Wirschaftsmathematik in Beruf und Ausbildung (Vol. 17, pp.
41-51). Wien: Holder-Pichler-Tempsky.
CC-STADUS. The Homepage of the CCSUSTAIN/CC-STADUS Project. Web-Address:
http://www.teleport.com/~sguthrie/cc-stadus. html.
Dormer, D. (1986). Die Logik des Miflingens. Strategisches Denken in komplexen
Situationen. Reinbek: Rowohlt.
Dorner, D. (1996). The Logic of Failure. New Y ork: Metropolitan Books / Henry Holt.
Educational Testing Service Network. . STACIN -Systems Thinking and Curriculum
Innovation Network Project. See: http://etsis1 .ets.org/research/stacin.html.
Forrester, J. (1961). Industrial Dynamics. Cambridge, Mass.: The MIT Press.
Forrester, J. W. (1992). System Dynamics and Learner-Centered-Learning in Kindergarten
through 12th Grade Education (D-4337). Cambridge, Mass.: Massachusetts
Institute of Technology.
Forrester, J. W. (1994). System Dynamics, Systems Thinking and Soft OR. System Dynamics
Review, 10(2).
Glasersfeld, E. v. (1995). Radical Constructivism: A way of Knowing and Learning. London:
Falmer.
Klieme, E., & Maichle, U. (1991). Erprobung eines Modellbildungssystems im Unterricht.
Bericht ttber eine Pilotstudie zur Unterrichtsevaluation . Bonn: Institut fiir
Test- und Begabungsforschung.
Klieme, E., & Maichle, U. (1994). Modellbildung und Simulation im Unterricht der
Sekundarstufe I. Auswertungen von Unterrichtsversuchen mit dem Modell-
bildungssystem MODUS. Bonn: Institut fiir Bildungsforschung.
Klir, G. (1991). Facets of Systems Science. New Y ork: Plenum Press.
Mandinach, E. (1989). Model-Building and the Use of Computer Simulation of Dynamic
Systems. J. Educational Computing Research, 5(2), 221-243.
Meadows, D. H., Meadows, D. L., Randers, J., & Behrens, W. W. I. (1972). The Limits to
Growth. New Y ork: University Books.
Ossimitz, G. System Dynamics and Systems Thinking Mega Link List. http://go just.to/sd.
Ossimitz, G. (1994). Endbericht zum Projekt "Systemdynamiksoftware im Unterricht"
(Manuskript ). Klagenfurt: Universitat Klagenfurt.
Ossimitz, G. (1996). Projekt "Entwicklung vernetzten Denkens" (Endbericht an die For-
schungskommission ). Klagenfurt: Universitat Klagenfurt.
Ossimitz, G. (2000). Untersuchungen zur Entwicklung systemischen Denkens. Minchen:
Profil.
Richardson, G. P. (1991). Feedback Thought in Social Science and Systems Theory. Philadel-
phia: University of Pennsylvania Press.
Richmond, B., Peterson, S., & Vescusco, P. (1987). An Academic Users Guide to STELLA .
Lyme, NJ: High Performance Systems, Inc.
Richmond, B. (1991). Systems Thinking: Four Key Questions . Lyme, NH: High Performance
Systems Inc.
Richmond, B. (1993). Systems thinking: critical thinking skills for the 1990s and beyond.
System Dynamics Review, 9(2), 113-133.
Richmond, B. (1994). System Dynamics/Systems Thinking: Let's just Get On With It
(Conference Paper ). Lyme, NH: High Performance Systems, Inc.
Schloglhofer, F. (1993). "Das Lehrplankapitel "Untersuchung vernetzter Systeme" im
Osterreichischen Mathematiklehrplan der Oberstufe der AHS. Universitat
Klagenfurt.
Vester, F. (1986). Unsere Welt - ein vernetztes System. Miinchen: Deutscher Taschenbuch
Verlag.
Walser, W., & Wedekind, J. (1991). Konzeption eines grafisch interaktiven Modellbildungs-
systems (Arbeitsherichte Medienforschung 4). Tibingen: Deutsches Institut fiir
Fernstudien.
Waters Foundation (ed). (1996). Systems thinking and System Dynamics in K-12 Education :
Waters Foundation.
Wedekind, J., & Walser, W. (Eds.). (1992). Modellbildungssysteme - Konzepte und Realisier-
ungen. Duisburg: CoMet Verlag.
* VUDYNAMO is an easy-to use DYNAMO-clone for PC computers, suitable for the early XT
and AT generations with primitive CGA-Graphics. VUDYNAMO accepts the original
DYNAMO syntax, but does not require the (j, k, 1) time indices of the original Dy NAMO.
* The Gymnasium is a non-business-oriented type of secondary school with a lower school
from grade 5-8 and an upper school from grade 9 - 12. It is finished with the Matura/A bitur
leaving examination, which permits general University access in Austria. The
Realgymnasium is a branch of the Gymnasium with a somewhat higher natural science
orientation.
