Systems Thinking &

Systems Modelling

A Course for Understanding Systems and Creating Systems Models

Systems Thinking and Systems Modelling Course

These materials were created as part of The Sustainability Laboratory’s educational

curriculum. It was developed with financial support from The Foundation for Advancement of
Systems-Orientated Management Education at the University of St. Gallen, Switzerland.

The introductory module, Understanding Systems, was written by Dr. Michael Ben-Eli, the
systems modelling modules were developed in collaboration with Loops Consulting and this
e-book version was created by Systems Innovation.

An online version of the course is also available on the Kumu platform:

Click here to find the full online course
Introducion
By Michael Ben-Eli, founder of The Sustainability Laboratory

The sustainability challenges we face on the planet are multifaceted, complex, and
interconnected. To understand them and address them effectively requires a holistic
“systems” view that is still largely absent from most mainstream approaches in science,
government, business, and education. It could be argued, in fact, that the many
sustainability-related issues—climate change, desertification, potable water shortage,
biodiversity loss, economic and social instability, and more—are the result of a non-systemic,
fragmented, simplistic, and short-sighted world view that dominates our industrial civilization.
Recent decades, however, have seen a significant surge of interest in holistic ways of
looking at reality with the associated development of multiple frameworks and tools which, all
together, have been hailed as the emergence of a new paradigm. The combined result has
been referred to in the related, burgeoning literature as “systems thinking,” “the system
approach,” or “the system view of the world.”

Purpose of This Course

At The Sustainability Laboratory, we subscribe to the view that systems thinking and a
deeply ingrained ability to “think systems” is an indispensable tool for future sustainability
leaders. We have therefore made systems thinking and systems modelling a central part of
The Lab’s educational curriculum. The materials that follow constitute a key element in our
Global Sustainability Fellows (GSF) program. It combines a number of teaching modules that
divide logically in three essential parts: an introductory overview to systems thinking; the
fundamentals of systems mapping and systems modelling; and a selection of illustrative
case studies.

This short but comprehensive “course” has been designed to blend theoretical, conceptual
exposure with hands-on engagement and collective building of systems models in groups.
The course has been developed in collaboration with members of Loops Consulting, led by
three of its leaders who, themselves, are alumni of the GSF program. The work was made
possible by a grant from the Swiss, University-of-St. Gallen-based, Foundation for the
Advancement of Systems-Oriented Management Education. Board members of the
Foundation are, in their own right, leaders in the systems thinking movement.
Overview of the book
This course is divided into five modules.

Module One gives an introduction to the field of systems thinking, discussing its emergence
as a new scientific paradigm and elaborating on some of the characteristics of a "systems"
view of the world.

Module Two introduces the field of system dynamics, which is itself a methodology to help
us think about issues from a systemic perspective. Here we discuss why we build models,
and how they can help us to think and communicate more deeply regarding certain issues.
Note that in this course we will deal only with qualitative system dynamics models, known as
causal loop diagrams (CLDs), and will not be dealing with quantitative simulation models.

Module Three introduces the reader to the basic building blocks of CLDs, explaining how
they use named variables and arrows to display the different cause and effect relationships
that are assumed to give rise to a certain patterns of behaviour.

In Module Four​, you will get the chance to apply the set of skills you acquired from the
earlier modules, as we lead you through our own thought process for creating CLDs about
selected issues. Here we will also introduce the concepts of stocks and flows, which are
important in the field of system dynamics.

In the fifth and final module, we will discuss five case studies that demonstrate how CLDs
have been used in particular situations, and the value that they provided.

The ultimate goal is that by end of this course, you should be able to understand and even
build your own causal loop diagrams, as well as have a better grasp of what it means to
think in systems.
Module One: Understanding Systems
By Michael Ben-Eli

YELLOW-RED-BLUE BY WASSILY KANDINSKY

IMAGE COURTESY OF GUILLAUME PIOLLE VIA WIKIMEDIA COMMONS
A Personal Note
The story of the essential link between issues of sustainability and system thinking is intensely
personal for me. It began in the 1960s, when as a young architectural student in London, I met
Buckminster Fuller. This life-changing event launched a long association with him first as a student,
and then as a collaborator in several projects in different parts of the world. This fruitful
association opened my eyes to many of the issues addressed today under the “sustainability
umbrella.”
An intuitive recognition of the essentially systemic nature of such issues drove me to explore
concepts that were emerging at the time from the system sciences, and these included system
dynamics courses at MIT. In the course of a management consulting career, opportunities soon
developed to explore the use of systems dynamics as a tool for synthesizing strategies in different
sectors and in relation to different issues, and a few of these cases will be touched upon further
below.
This preoccupation with a system perspective then led to pursuing a doctorate in cybernetics,
studying with some of the leaders in this discipline. Along the way, as a student, co-worker, or
friend, I have had the fortune to encounter some of the key protagonists in the field, including
colleagues and friends from the St. Gallen school.
Purpose of this Module
The purpose of this module is to provide a basic familiarity with the concept of “system.” We shall
touch briefly upon the place of the concept of systems in science, review definitions of the term
“system,” and highlight the essential characteristics of systems. We shall briefly encounter general
system theory and cybernetics as two main complementary scientific disciplines that underlie
systems thinking, make the link to system dynamics as a rigorous tool for engagement with
systems, illustrate the qualitative significance of the use of this tool in a couple of instances of
synthesizing management strategy and, finally, speculate on the implication of adopting the
system view of the world.
The intention is to take a random, introductory walk in the systems landscape, present the flavor
of key concepts, trace their origin, and emphasize their significance. The material follows no
particular order or strict linear logic, but rather mimics the very nature of “systems” by offering a
smorgasbord of ideas that, with some overlap, linkages, and perhaps even some repetition, reflect
the essence of a whole.
The table of contents on the following page gives an overview of the concepts discussed in this
module.
Contents

1. The Concept of System in Science......................................................................................................... 1

2. Definition of “System”........................................................................................................................... 3
3. Classification of Systems ....................................................................................................................... 5
4. The Essential Characteristics of a System.............................................................................................. 6
5. General System Theory ......................................................................................................................... 7
6. Organization as Organism ..................................................................................................................... 8
7. Structures and Mechanisms: The Cybernetic Perspective .................................................................. 10
8. Measuring Complexity ........................................................................................................................ 13
9. Other Key Cybernetic Terms ............................................................................................................... 14
Synergy ........................................................................................................................................... 14
Self-Organization ............................................................................................................................ 14
The Law of Requisite Variety .......................................................................................................... 15
Redundancy of Potential Command ............................................................................................... 15
Ultrastability ................................................................................................................................... 16
Reflexivity ....................................................................................................................................... 16
10. System Dynamics as a Tool for Engagement ....................................................................................... 17
11. Some Illustrative Cases........................................................................................................................ 18
Health Care..................................................................................................................................... 18
Great Lakes Water Level Fluctuations ............................................................................................ 21
Nuclear Waste Management ......................................................................................................... 24
12. Ethical and Practical Implications of the Systems Approach ............................................................... 25
Further Readings ......................................................................................................................................... 28
 1. The Concept of System in Science
Imagine an organism of some kind that is struggling to survive, perhaps even thrive, in its world.
Its prospects would be greatly enhanced if it were able to utilize an accurate map of its
environment—a set of explanatory models of its surroundings—to guide its actions. Such models
can take many forms, from a simple image, to a parable and a story, to a complex scientific theory.
From the earliest dawn of conscious awareness, human have been creating implicit or explicit
theories about the world in order to navigate its challenges and complexities. Whether such
theories are primitive or sophisticated, ancient or current, they all ultimately depend on the
accumulation of experience by trial and error. For example, by repeatedly eating berries and
experiencing the effects, humans slowly began to learn which were lethal and which were good
for their health. By putting a new scientific theory to experimental test, it is established whether
the theory works or if it must be discarded and replaced.
What has changed along the long and arduous trajectory of the evolution of knowledge are the
essential characteristics of methods used in the construction of explanatory models. Changes in
the central features of such methods are commonly identified by philosophers of science as major
historical milestones, profound shifts in paradigms that shaped not only the nature of acquisition
of knowledge, but also affected the very nature of social and economic reality.
For thousands of years, explanatory models—basically superstitions with strong animistic
flavors—were associated with “forces” external to the phenomena that they were meant to
explain. Things happened, and events unfolded as they did, because of the intervention of external
factors: the spirits, the gods, mysterious arbitrary forces, and the like.
In the Western tradition, the early Greek philosophers are credited with bringing about a
fundamental shift, a shift that is relatively recent—starting some two and a half millennia ago—
which laid the foundations for what was later to emerge as the modern method of science. Greek
philosophers laid the basis for experience-based explanation. They introduced the method of
orderly observation and documentation of various aspects of nature. They also established the
practice of seeking explanations by decomposing elements of the world to their component parts,
down to the elemental building blocks, the “atoms,” then seeking to reconstruct an understanding
of the whole. The significant shift then is in moving from an explanation by an arbitrary external
cause to one that puts an emphasis on analytical decomposition of phenomena to their
component parts.
The next step in the trajectory, which brought about the advent of modern science, continued to
expand on the analytical, reductionist approach. This led to the dominance of the next important
shift: that of viewing the world as a machine. This development is associated with such names as
Copernicus, Kepler, Galileo, Bacon, and Descartes. It culminated with Newton’s mechanics and a
view of the world as a giant, clock-work-like, exquisite machine, obeying the simple laws of motion.

1
The scientific method itself established a rigorous procedure for the acquisition and verification of
knowledge. Observations are translated into a theoretical hypothesis, which is then subject to
experimental testing. Only when repeatable test results are consistent with the hypothesis, the
latter is accepted as an established theory. Underlying the procedure is the understanding that all
theories are essentially tentative and must be replaced when new observations contradict an
accepted truth.
The classical analytic approach has been responsible for a staggering and astonishingly rapid
expansion in our understanding of material reality. It has also been responsible for all the gifts—
and ills—of modern technology. By the 20th century, however, it began to run into considerable
difficulties, even in the physical sciences. It was in biology, in particular, that the classical model
proved completely inadequate in explaining the behavior of whole organisms by focusing on their
component parts. This ‘crisis’ in scientific theory, and the realization that many aspects of the
world cannot be explained by a reductionist method, catalyzed the emergence of the system
sciences. Emphasis was placed on highlighting patterns of interactions and interdependencies of
parts for the understanding of the behavior of wholes, and the concept of “system” was brought
to the fore.
In retrospect, we can discern a historical trajectory of approaches for explaining reality. The
trajectory begins with “animism,” with emphasis on arbitrary external forces; followed by
“atomism,” then “mechanism,” both exemplifying the classical, analytical, reductionist,
mechanistic world view; and then the concept of “organism,” a way of seeing the world as a
complex of relations, interactions and interdependencies. The latter constitutes the systems view
of the world.
Note that milestones along this trajectory are not as rigidly sequential as is implied by the listed
steps. They overlap and coexist as a mix. Also, throughout history, there were inspired individuals
who could see beyond the visibly obvious, seeking and expressing a holistic, integrated view of
reality. Think of Leonardo da Vinci, Spinoza, Blake, Hegel, and many more. Genuine mystics of all
times and traditions are also a good case in point.
Emergence of the systems view of the world has been hailed as a major revolution in scientific
thinking that was bound to replace the older mechanistic model. It might be useful, however, to
regard both as complementary theories that, existing in parallel, will continue to offer useful tools
for addressing different aspects of reality.
From the sustainability perspective, critical issues occur in a context that is exceedingly complex,
constituting a dynamic, multivariable universe that involves multiple social, political, economic,
and cultural aspects interacting among themselves, across sectors, and with the physical and non-
human living parts of the world. This systemic reality is irreducible, and it will not yield to a simple,
linear, analytic approach. Using reductionist tactics when dealing with systems is fundamentally
flawed, yet it is common when conceptualizing and attempting to address sustainability issues.

2
Approaching complex systems as though they were simple, clock-like mechanisms does not work,
and typically only exacerbates the very adverse conditions meant to be resolved.

2. Definition of “System”
What is a system?
The term “system” is often used in ways that miss the deeper connotations that are at the heart
of the relatively recent emergence of the systems perspective in science. The term is commonly
used to denote a concept of a totality, identified by some logical consistency, but without putting
an emphasis on the structure of relations, attributes, and causal interdependencies of the parts.
In this sense we speak about a system of law, a political system, an educational system, a health
care system, a production system, an economic system, and the like. These uses miss a deeper
point, and it might be useful, therefore, to examine a few select definitions found in the system
thinking literature.
The typical definition stresses the notion of system as any entity, any totality, consisting of
interacting parts. A straightforward, clear definition that was offered by Ludwig von Bertalanffy, a
German biologist and the father of general system theory, reads: “A system is a complex of
elements in mutual interaction.” This definition is free from all but one important criterion, that
of “parts in interaction.” The definition, incidentally, encompasses two important ideas. First, it
distinguishes between parts and wholes, a major topic in mathematical logic in its own right. And
second, it emphasizes the interaction of parts which make for the whole. Depending on the level
of resolution, any individual component can itself be regarded as a whole consisting of interacting
component parts. The focus on dynamic interactions and interdependencies makes all the
difference. It defines the shift introduced by the system view of the world.
Russell Ackoff, another systems pioneer, follows a similar line of thought, but introduces an
interesting twist by suggesting that a system can also be a conceptual entity. According to Ackoff,
a system is “an entity, conceptual or physical, which consists of interdependent parts.” This
observation expands the domain of possible systems and it has implications for the construction
of systems models, a topic that will be developed further below. For the moment, think about the
following: any set of interactions defines a specific set among all possible sets of relations. In this
sense, any given pattern of interactions represents a limiting factor that produces a particular kind
of behavior and no other.
The next definition raises the level of abstraction by highlighting the dynamic consequence of
systemic interactions. It emerged, in the late 1960s, out of the Information System Theory Project
(ISTP), which was organized to address the description of data structures. It defines a system as “a
set of mutually constrained events,” a powerful, compact definition indeed.
Another more pragmatic approach is taken by Jay Forrester, the originator of system dynamics,
the modelling methodology originally developed at MIT. Forrester defined a system as “a grouping
of parts that operate together for a common purpose.” Inevitably, this definition introduces

3
questions about how a “purpose” is defined. It highlights the role of an observer in interpreting an
examined phenomenon. An observer can assume a purpose, for example, by suggesting that “the
purpose of bees is to cross-pollinate,” or perform the role of a creative agent who actually
introduces a new purpose into the world, for example, in declaring "the purpose of my invention
is to purify water”.
A definition that comes from an entirely different tradition is due to Buckminster Fuller, the
futurist architect-designer. In his seminal work Synergetics—Explorations in the Geometry of
Thinking, Fuller defines a system as “the first subdivision of universe into a conceivable entity.”
Such a subdivision is an act of cognition and, again, the role of an observer comes to the fore. It is
the observer who subdivides total reality in an act that separates the focus of attention from all
the rest.
Finally, a minimalist, elegant definition that is associated with information theory and the work of
Ross Ashby, states simply that “a system is a source of information.” This definition may obscure
the emphasis on parts and interactions, but like other rigorous definitions, it embodies fascinating
philosophical and practical questions. These are associated with profound issues related to our
deepest concepts about the nature of reality, purpose, identity, cognition, the nature of
knowledge, the role of intuition, and the place of consciousness in the world.
Raising the question of the role of cognition and anchoring the identification of a system to a
mental act suggest that establishing a system’s identity is an act of intelligence imposing its own
criteria on the world. These criteria force practical but arbitrary separation with the consequent
fragmentation of a reality that ultimately comprises one system—the meta system of all systems.
This is, of course, what profound spiritual visionaries have been telling us all along.
A couple of additional observations that relate to the role of cognition might be useful for the
aspiring systems modeler. These have to do with the challenge of selecting the right boundaries
when modelling a system. This question of appropriate boundaries is ultimately dictated by a
particular purpose, and it depends, to a large extent, on the modeler’s experience and skills. In
addition, as a product of a mental act, every system’s model will always remain an abstract
representation of reality. A model, it should never be forgotten, is not the real thing. It is a
pragmatic device, a necessary simplification of reality that helps make the issues we need to
address in a complex world handleable. It can offer only an approximation and no matter how
elaborate, it will never capture the full richness of the real world.
Finally, most consequential systems that we interact with are dynamic in their very nature: they
continuously change, adapt and evolve. This dynamic aspect is captured by the idea of the self-
organizing system, which requires that observers continuously change their frame of reference
and, often, their conclusions. There is a “time-sensitive” element involved here, suggesting the
virtue of keeping an open mind and perhaps even fostering a measure of playful flexibility when
dealing with a complex world. Remember Shakespeare? “There are more things to heaven and
earth, than is dreamt of in your philosophy…” Contemplate for a moment what this statement

4
means to the effectiveness of adopting stubborn, rigid dogmas when dealing with the many
interlocking ecosystems that make our world.

3. Classification of Systems
Following the emergence of general system theory, several attempts have been made at
developing a classification of systems. Classifications were offered based on categories such as
function, sector, purpose, and other conventional attributes. These efforts have not contributed
decisively to a better understanding of systems; instead they essentially reflected the existing
frameworks of scientific and professional disciplines. Definitive, rigorous classification has proven
difficult, perhaps due to the large variety and complex behavior of dynamic systems and the
challenge of highlighting the general aspects that transcend their specific embodiments.
A very useful classification, nevertheless, was offered by the prominent British cybernetician and
management scientist, Stafford Beer. The framework he proposed is helpful in honing attention
on the domain of phenomena, of exactly the kind of systems that cannot be addressed by the
reductionist, analytical approach.
Beer’s classification is based on a matrix consisting of two axes, as depicted by the figure below.
One axis consists of a three-fold distinction of simple, complex, and exceedingly complex systems.
The other axis consists of a two-fold distinction between deterministic and probabilistic systems.

Figure 1: Beer’s matrix used to classify systems

This matrix produces a distinction between six essentially different types of systems, which
ultimately vary by the number of relevant components and the multiplicity and effects of
interactions that are involved.

5
Deterministic systems represent domains where complete, reliable predictions are possible in
principle. First come the “simple and deterministic,” as in the case of a simple mechanical device;
next is the “complex and deterministic,” for example, an elaborate production system in industry,
other kinds of intricate technologies, or even a whole solar system, the gross behavior of which
obeys the classical laws of motion; next is “exceedingly complex and deterministic,” think of the
very complex system required for a successful lunar exploration and other space missions, where
thousands of people, extreme technologies, and millions of individual tasks are involved, and still
strict control must be assumed.
Under the probabilistic criteria, in turn, we find the cases where complete predictions are not
possible and only various degrees of likelihood of occurrences can be discussed. First, the “simple
and probabilistic,” for example, a coin throw; next, the “complex and probabilistic,” the behavior
of stock markets offers a good example; and, finally, we converge on “the exceedingly complex
and probabilistic,” the domain pertaining to the kind of highly dynamic systems, multiple
components and multiple interactions, numerous interdependencies, variable and unpredictable
behaviors, and irreducibility in principle. This is the domain of living systems, including those of
the socio-economic variety.
Note that the language of this classification reflects the somewhat older language of operational
research and early cybernetics. Today, the category of exceedingly complex and probabilistic
systems would be identified with the realms of chaos and complexity theories. Note too that
sometimes categories overlap and that they do not always represent strict boundaries. A selfsame
system may have to be described differently at times, depending on an observer’s resolution level
and purpose.
The important point is that the seemingly cumbersome designation of the “exceedingly complex
and probabilistic” system, helps focus our vision on the fundamental characteristics of the very
realm that is essential for our ultimate wellbeing as humans: the realm associated with life. This
realm represents the broad ecosystem issues that as participants-designers we need to be
engaged with, in the quest of realizing the promise of a better world.

4. The Essential Characteristics of a System

Perhaps the most succinct formulation of the essential characteristics of a system is due to Russell
Ackoff, the leading American organizational theorist and system thinker. This formulation goes
right to the essence, by focusing on the nature of the interactions that account for the concept of
systems in the first place.
According to Ackoff, three essential conditions must be fulfilled in characterizing a system. These
include the following:
First, that in a system each simple element has an effect on the behavior of the whole.
Second, that in a system each element is affected by at least one other element, and none has an
independent effect on the whole.

6
Third, that in a system no subgrouping of elements into totally independent subsystems is
possible.
If you think of a visual depiction of a system as a network of dots representing key elements, with
arrows drawn between them to indicate the underlying interactions, then in such an arrangement,
by virtue of its links to other parts, each individual element is an integral part of the network,
contributing to the behavior of the whole. Each is connected by an arrow to at least another
element and each is linked, in turn, by an arrow originating from one other element, or more.
Finally, every conceivable subgroup of parts connected by lines among themselves is always arrow-
linked to the rest.
Whether applied to an organism, a rainforest, an urban complex, a business enterprise, or society,
this characterization is at the heart of the slowly emerging ecosystems perspective, which
highlights the vital significance of interdependencies. The philosophical as well as the practical
implications of internalizing this perspective are huge. Contemplate, for example, the implications
to the Judeo-Christian mythology of creation and its portrayal of a creator as an all-powerful force
acting independently and from outside the system that is brought into being; or think of the
arrogant leader, or manager, claiming full control, and acting as though they were “above” the
system that they are trying to affect.

5. General System Theory

It was in the field of biology that the limitations of the classical reductionist methods of science
were first emphasized, in a shift that established the system concept as a new paradigm. This shift
was led by Ludwig von Bertalanffy.
The impetus came from a dawning realization that living systems show essential qualities that add
up to more than the sum of their parts and depend on the integrity of the organism viewed as a
whole. The validity of this synergetic, non-linear effect, incidentally, is not confined to biology, and
it is relevant to aspects of the physical sciences as well. A simple manifestation can be found in
metallurgy and the creation of high-performance metal alloys. For example, specific properties of
individual molecules, such as tensile strength, yield much higher tensile performances in
combination with specific other molecules, than would be expected from a simple sum of the
corresponding parts.
Bertalanffy showed that the effectiveness of the reductionist analytical procedure depends on two
key conditions. First, that interactions between parts are non-existent, or negligible, and second,
that it should be possible to simply add up descriptions of parts to construct a picture and an
understanding of the whole. As a tool of inquiry, it became apparent that the reductionist method
was limited to dealing with two, or at most three, variables. It is not applicable to phenomena
characterized by multiple variables, complex interactions, and non-linear behavior. Using the
strictly analytical approach inherited from 19th century science, it was possible to tell how one cell,
or one organ, reacts to one particular stimulus or how one entity reacts to the application of one

7
kind of force. It was not possible to tell how several different variables act together when exposed
to a number of different influences at the same time.
Bertalanffy advocated the importance of an “organismic” perspective, and the combined focus on
patterns of interactions and the notion of systemic wholeness led to a new emphasis: the concept
of “organization” as the basic unit of study. The consequence was a view of organisms as
organizations and the invitation to regard organizations as organisms. The emergence of the idea
of organization with its related stress on an underlying structure had a profound impact. In biology,
an integrated concept of organization as a new principle replaced the need to rely on the idea of
a “vital force” in explaining the special qualities of living systems. It also led to a better
understanding of the relationships between specific structures and associated behaviors.
Understanding the basic principles of organization drove the new way of inquiry, and the search
for such principles became the central goal of general system theory. The quest aimed to produce
a unifying theoretical framework across all systems in general.

6. Organization as Organism
As commonly used, the term “organization” carries the connotation of a discernable order and it
is usually applied in reference to some entity, typically with institutional overtones, such as a
business, government, party, or any other social or economic body.
The emphasis put by general system theory on the concept of organization as a new defining
paradigm shifted the meaning to another kind of abstraction. The shift stressed the significance of
underlying structures to the representation and understanding of a system’s behavior. This idea,
as we shall see, had significant implications, especially to developments in cybernetics, a closely
related field.
The point is that a particular structure is embodied in sets of relations. In the systems thinking
context, a set of elements form an organization when some specific relation defines their
interaction—when a specific pattern of relations is being conserved. A specific pattern of relations
acts as a constraint, limiting the number of conceivable configurations and reducing the field of
possibilities to the unique one that is then manifested as a distinct organization.
The “nesting,” recursive property of systems, whereby at a higher level of resolution, any element
in a system represents a whole other system in its own right, introduces the idea of hierarchy. This
nesting quality is akin to a sequence of views, as in satellite pictures, where one can zoom in on
increasing levels of detail. A view of Earth, for example, is resolved into an image of a regional
geography, then to that region’s main features, such as mountains, forests, rivers, and lakes, and
all the way down to the minutest, most-specific details.
In the system thinking context, the concept of hierarchy found its expression in a view of the world
as a stratified organization of increasing levels of complexity, stretching all the way from
elementary forms of matter to the highest manifestations of life. The idea is often expressed as a
sequence: basic particles, atoms, molecules, various forms of matter, simple life-forms, complex

8
organisms, whole ecosystems—including human society—and so on. Note that levels in such a
hierarchy corresponded to levels of emergence of novel qualitative properties, which distinguish
one level from the next.
Each level represents a cluster of interacting sub-components, and levels can be distinguished by
the relative strength of respective interactions. These interactions would be stronger within each
level and weaker between levels. It is the relative strength of such relations—internal bonds, if
you will—which allows for a definition of boundary conditions and makes the individual integrity
of a level stand out against the background of its environment. A fascinating question in this regard
is whether such a logical stratification reflects an essential property of the world and is inherent
to the process of evolution, or whether it is the result of the structure of language, cognition, and
the nervous system. It may turn out that the very distinction—cognition-language-nervous
system-the world—is itself arbitrary, and that we are actually facing here a deeper affinity, a basic
manifestation of one, continuous, co-generating, co-defining reality.
A particularly significant illustration of the aspect of nestability, or verticality, is borne by Fuller’s
definition of a system. Recall that Fuller defined a system as “the first subdivision of universe into
a conceivable entity.” Fuller’s perspective is anchored in geometry and from that view point a “first
subdivision” is represented by the simplest possible three-dimensional configuration, a minimum
set of relations that represents a stable structure. This minimal set corresponds to a tetrahedron,
a four-sided, triangular-faced pyramid.
Imagine such a structure with its four vertices, four faces and six edges. As Fuller pointed out, it
subdivides the world into all that is outside the structure, the structure itself, and its interior. This,
of course, offers an immediate, visual illustration of a three-tier hierarchy. Upon first glance, this
may sound like an esoteric abstraction. However, it has profound, practical implications for all
system designers. It means that for any constructive intervention, one needs to optimize a
coherent, harmonious integration of the relevant aspects of at least three levels: the system under
consideration, its environment, and its internal components. This is another way, incidentally, of
addressing the question of how to set system boundaries when analyzing a particular systemic
challenge or synthesizing a preferred system configuration.
In the realm of living organizations that are characterized by numerous variables and multiple
interactions, the organizational paradigm calls for adjusting static concepts to a dynamic reality.
The significance of this essential dynamism cannot be overemphasized. It is the inevitable
consequence of the idea of “interaction” itself. At each case, groups of elements interact among
themselves and across levels, settling into temporarily stable configurations—self-preserving
forms—that are adapted to the constraints of their specific context. This brings us back to the
concept of self-organization. A self-organizing system is an “open” system that is engaged in a
constant metabolic exchange with its world. It imports substances and releases the by-products
of its activity to an environment that is populated by many other species of equally active self-
organizing systems. The constant mutual adjustments, adaptations, and occasional
reconfiguration of novel states of equilibrium require that an observer—a manager,

9
experimenter, or an agent of change—stays alert. It means remaining open to adjusting frames of
reference, and adapting assumptions, decisions, and actions to accommodate the reality of a
changing world.
From this perspective, the universe can be regarded as an immense, kaleidoscopic flux of
constantly interacting, inter-transforming, sub-system events. “Reality” is forever dynamic, and
distinctions that identify its individual elements are introduced by changing resolution levels
imposed by observers. In this broader context, management, planning, design, and other forms of
proactive intervention-type activities are among the processes through which self-organization is
manifest in the social domain. An urgent challenge of our time is how to design agile, responsive,
adaptive, intelligent, and inclusive organizations with the capacity to deliver an enduring
advantage for all. This is especially the case, since all evidence suggests that most of the dominant
socio-economic structures inherited from the past are no longer up to the task of securing a
sustainable trajectory in a flourishing world.

7. Structures and Mechanisms: The Cybernetic Perspective

General system theory laid claim to the space of seeking prototypical characteristics, generic
features, and underlying principles that hold true for systems in general and transcend specific
embodiments. In this context, cybernetics emerged as a closely related discipline focusing
specifically on understanding the dynamics of a system’s internal structures and the mechanisms
that maintain a dynamic organization invariant. Cybernetic theories put emphasis on questions of
how systems regulate themselves, how they adapt and evolve, how they self-organize and, more
specifically, what the structures and mechanisms are that mediate their operation, viability,
performance, and conduct.
The story of cybernetics is fascinating, and it merits a brief introduction. During the Second World
War, as the speed of aircrafts increased significantly, humans who maned anti-aircraft guns
became too sluggish in tracking a target, aiming at it, shooting, and correcting for deviation when
they missed. In response, a project was organized at MIT, under the leadership of Norbert Wiener,
a prominent professor of mathematics, to develop automatic control mechanisms for anti-aircraft
guns. This called for combining capabilities in tracking a moving body, predicting its future location,
targeting, and error controlling. Tackling the challenge required the integration of inputs from a
number of disciplines, and Wiener’s circle included engineers, mathematicians, physiologists,
neurophysiologists, information theorists, and early computer scientists. As discussions
proceeded, it became apparent that questions of communication and information were central to
all processes of control in general.
A conclusion emerged that the logic of circuitry involved in man-made, automatic, error-control
mechanisms was analogous to the logic of homeostatic structures found in physiology. These
structures work to maintain specific physiological values—body temperature, body fluid
composition, blood pressure, blood sugar level, and the like—by triggering an appropriate
corrective action when deviations occur from a norm. In all such cases, both mechanical and

10
biological, circular mechanisms of feedback nature are involved—mechanisms that correct for, or
amplify, differences from a given norm. These involve two essential types of loops: negative and
positive feedback loops, respectively.
Negative feedback loops are associated with goal seeking behavior, in which a difference between
an actual and a desired condition (hence, “negative”) is used as an input to bring the actual closer
to the desired value. Positive feedback loops are associated with self-reinforcing, self-amplifying
behavior, such as associated with processes of growth and decay, where a particular outcome
generates an even greater outcome of its kind. An important novelty was in linking the concept of
controlling or regulating processes to the results of a system’s actual performance. The concept
of purposeful behavior could now be anchored to specific structures, processes where information
content and flow play a central role.
The universality of the fundamental nature of regulating mechanisms, their applicability to the
functioning of systems in general, and the fact that one theoretical framework was valid to
describe regulation processes in all cases, were novel insights. This merited a new scientific
discipline, one that would address communication and control processes. In 1948, Wiener’s
classical book, Cybernetics, was published and a new science had been born.
In his book, Wiener defined cybernetics as “the science of control and communication in the
animal and the machine.” This definition highlights two key ideas. Firstly, it classifies “control” and
“communication” together, indicating the role of information in processes of regulation and
control. And secondly, it claims the validity of the theory to both man-made devices and living
systems. The term “cybernetics,” incidentally, denotes the role of feedback mechanisms in
processes of regulation and control. The term is derived from the Greek word for steersman, the
person steering a boat to its destination, correcting for the influence of currents and winds. At first
glance, “control” may appear to be a restricting concept. It is meant, however, in the most general
sense of regulation processes, processes which mediate for particular outcomes and bind a system
together as they preserve its singular identity. As it turns out, in complex dynamic systems, the
structures that drive ultimate behavior take the form of intricate networks of loop-like, circular
interactions, dominated by the familiar feedback mechanisms discussed above, that restrain or
amplify specific conditions.
Making the link between a system’s internal structure and its actual behavior was a huge step in
understanding the conduct of complex systems of all kind. The idea goes back to Wiener and his
colleagues, who in a seminal 1943 paper, Behavior, Purpose and Teleology, established the
indispensable connection between a system’s output—its observable behavior—and its internal
structure. Historically, this insight was profound. First, it clarified the question of purposive
behavior, and in tying a system’s behavior to specific internal structures, it removed, once and for
all, the need to resort to notions of “vitalism,” and similar vague explanations. Secondly, it made
it clear that in order to modify a system’s behavior, a change must be made to its core structure.
Simple as this may sound, think of how often, in trying to reform a system, efforts are directed at
the behavior itself, rather than at the structures that drive it.

11
The basic tenets of cybernetics and the notion of abstracting principles of regulation to a level of
comprehensive validity were taken significantly further by the British cybernetician Ross Ashby.
Early concepts in cybernetics were derived from observations and direct experience with man-
made servomechanisms and homeostatic processes in the body’s physiology. From these direct
experiences with actual systems, some general principles were derived. Ashby was able to extend
the vocabulary of cybernetics by turning the method around, working from the abstract and
general to sketch out a logic of regulation processes in general, which could then be related to
particular cases in technology, biology, or society. The idea was that cybernetic theory would thus
refer to regulation processes in all conceivable types of systems, in the same way that geometry,
for example, relates to all kinds of objects in space. Ashby’s seminal book, Design for a Brain,
remains a classical source for our understanding of the dynamic behavior of complex systems.
Concepts from cybernetics were soon utilized in a broad range of fields. For example, in research
concerning neural networks, in the work of Warren McCulloch and Walter Pitts; in anthropology,
with the work of Gregory Bateson, Margaret Mead, and Roy Rappaport; in relation to a theory of
learning, in the work of Gordon Pask; in cognitive psychology, in the work of Paul Watzlawick and
his colleagues; in relation to a theory of self-organization, cognition, and the role of observers in
constructing reality, with the work of Heinz Von Foerster; in management, with the work of
Stafford Beer; in developing system dynamics modelling, with the work of Jay Forrester; and more.
An instructive example of importing concepts from cybernetics to the study of social systems is
provided by the work of anthropologist Roy Rappaport, at the time at Columbia University. The
subject of the study conducted during the 1960s was a small, aboriginal community of hunters
and gatherers, the Tsembaga, who inhabited a remote territory within the Bismarck mountains of
New Guinea. The group had been only minimally exposed to outside civilizations and could thus
offer a good case study of a well-defined, local community that was well-adapted to the specific
circumstances of its environment. Life of the Tsembaga was regulated by a ritual cycle—a
prescribed routine—involving various activities and festivals performed over several years. The
study was able to show how the ritual cycle acted as a homeostat, regulating the relationships
between members of the group, and between the community and key elements in its
environment, which together comprised a whole ecosystem.
In a nutshell, the Tsembaga lived in a close, symbiotic relationship with domesticated pigs that
they raised with great care. Pigs provided an important source of proteins in times of stress. They
assisted in cultivating gardens by digging for roots, eliminating weeds, and turning and softening
the ground. They also kept residential areas clean by feeding on waste. Pigs represented the most
important part of the group’s wealth. When demands of the combined human and pig population
exceed the carrying capacity of the group’s territory, a year-long festival was launched during
which pigs were slaughtered, meat was shared, the size of the herd was drastically reduced, and
balance was restored. The ritual cycle involved many other aspects, regulating inter-tribal relations
and regional social contacts, controlling warfare, and more. This fascinating story is recounted in
Rappaport’s book Pigs for the Ancestors. The crux of the matter is that the ritual cycle was shown

12
to maintain essential variables within “desired” limits, by triggering actions that restore the system
to desired norms when threatening deviations occur.
The story of the Tsembaga describes the case of an established adaptation and a socio-cultural
mechanism operating to maintain a prescribed steady state. It does not address the circumstance
of a major change in a context that would require a complete systemic shift. Nevertheless, it is
pertinent to current issues of sustainability because it highlights the question of the interaction
between a population and the carrying capacity of its environment, precisely the kind of question
we face on the planet today.

8. Measuring Complexity
We talked about complexity as an inherent characteristic of dynamic, “living” systems. But what,
more precisely, does the term “complexity” mean? The idea of complexity is often confused with
size: the bigger, the more complex. This is an error, of course, since even a tiny, unicellular
organism represents a highly complex system. Conversely, adding ever more grains to a heap of
sand does not make it more complex. Ultimately, the idea of complexity is independent of size;
rather, complexity relates to the number of possible distinctions that can be made about a given
system.
In cybernetic terms, possible distinctions relate to a system’s internal variety, where “variety”
expresses the number of a system’s possible states. The higher the number of different elements
and the higher the number of pathways through which these elements interact, the higher the
system’s variety. A simple electrical light switch with only on or off distinctions has a relatively low
variety. If a dimmer is used to control light levels, it has a higher range of possibilities
corresponding to a higher variety. A biological organism, a society, an economy, or an ecosystem
all represent systems of exceedingly high variety.
Complexity is thus measured by a system’s potential variety: the number of different states that a
system can assume. In dynamic systems, variety can proliferate very rapidly. Think about it this
way: with large numbers of elements, a high number of interactions, and the possibility of each
interaction assuming more than a single value, the number of possible states can increase
exponentially and quickly.
There is an obvious sense in which the measure of a system’s variety represents an observer’s
discrimination capacity. It also represents an observer’s level of uncertainty, in that the total
number of distinguishable states of a system signifies the observer’s uncertainty about it—which
of the many possible states will be assumed next? A quantity of variety thus offers a measure of
uncertainty as well.
This is profound. Uncertainty relates to our perception of order, which relates, in turn, to
detectable regularities in a system’s behavior—a sequence of a system’s changing states. Such
regularities, incidentally, are produced by constraints imposed on a system’s potential variety by
its internal structure. Uncertainty relates, accordingly, to a quantity of entropy, a measure of

13
disorder in a system. Because the appearance of one state out of all possible states of a system
removes some uncertainty, in the process conveying an amount of information, the concepts of
uncertainty and information are closely related. They assume a similar mathematical expression,
yet have opposite signs. For example, when uncertainty is at a maximum, which happens when all
events in a given universe may occur with equal probability, there is no information available and
variety is naught. The concept of variety is thus at the nexus of ideas involving physics, information
theory and the philosophy of science, and linking concepts of organization, entropy, and order, all
of which are key to our understanding of the world.
A thorough appreciation of the concept of variety and the related ramifications is vital for the
system practitioner. Designing novel system configurations in the real world involves the art of
mastering the intuition for when to amplify and when to attenuate variety, a typical task in
managing any enterprise, and the essence of regulation in general. The concept also suggests the
level of respect, and humility perhaps, that should be adopted with any design intervention in the
socio-economic and eco-systemic space.

9. Other Key Cybernetic Terms

A selection of a few key cybernetic terms merits a brief exploration. All relate to one another, as
well as to the concept of non-linearity that is at their base. They offer valuable tools for efforts
aimed at system synthesis and design.
Synergy
We have already encountered the term synergy. It pertains to the behavior of whole systems,
which is unpredictable by the behavior of the parts. Synergetic effects account for the
emergence of new, often unexpected properties. The inevitable expression of synergetic
effects in social interactions ought to be assumed and incorporated in all cases of social-
systems design. Synergetic effects can be a source of unwelcomed surprises when they
produce and amplify unintended consequences. At the same time, synergetic potentials can
be deliberately used to create and increase added value to desired results.
Self-Organization
The concept of self-organization dominated early discussions of cybernetic theory. An elegant
definition is due to Von Foerster. It is based on the concept of “redundancy,” imported from
information theory. In information theory, the term redundancy is used in the context of
protecting information integrity from deterioration due to effects of background noise, by
augmenting information content or increasing channel capacity. Simply stated, redundancy
allows for more potential “possibilities.” According to Von Foerster, a system is self-organizing
if the rate of change of its redundancy remains positive.
This formulation is strongly linked to the concepts of order and order creation. Redundancy
will be nil at a state of maximum disorder, when no distinctions can be made and activity
ceases. On the other hand, a system can continue to exhibit dynamic, self-organizing

14
characteristics—maintaining or increasing its manifestation of organization and order—as long
as a sufficient level of redundancy and the system’s internal complexity are preserved.
The dynamic characteristics of self-organizing systems have already been alluded to in
previous sections. These properties force an observer who interacts with a self-organizing
system—or who is a part of it—to keep an open, dynamic stance. How to work in tandem with
the self-organizing properties of a system, rather than destroy them, is the challenge of all
interventions in the socio-ecosystemic domain.
The Law of Requisite Variety
The law of requisite variety is due to Ashby. It has emerged as one of the central tenets of
cybernetics and is fundamental to the general theory of regulation and control. Ashby’s law
states that “only variety can absorb variety.” Effective regulation can only be achieved when
the regulating system contains at least the same amount of variety as the system being
regulated. The requirement for requisite variety is applicable to all systems: automated
devices, technology processes, ecosystems, and social systems alike.
The law of requisite variety may sound very simple. It looks obvious once it is recognized. But
contemplate the typical way that most organizations in business, government, and other
aspects of human affairs are managed and structured. More often than not the conventionally
familiar organizational structure perpetuates a management model that imposes structures
with grossly insufficient variety, a direct consequence of a non-system-based, reductionist
world view. Such a management model is simply inadequate―not rich enough―to address
the demands of an increasingly more complex world. Among corporate leaders none
understood this better, perhaps, than Dee Hock, founder and CEO emeritus of VISA
International. Over a long and fruitful career, Dee Hock highlighted the gross deficiencies of
the prevailing, hierarchical, command and control structure. He advocated a more versatile
and dynamic form of organization and coined the term “chaordic” to describe it. The term
itself suggests a creative combination of “order” and “chaos”, which is at the heart of the ability
to experiment and innovate.
Redundancy of Potential Command
This term was coined by Warren McCulloch and his colleagues, in relation to research done on
the workings of neural networks and, in particular, the reticular formation. The reticular
formation is a network of neurons in the brainstem that regulate various aspects of behavior,
including states of consciousness.
The research found that such networks function effectively by virtue of their enormous
redundancy, born by the huge number of individual nerve cells and the multiple paths of their
interactions. A typical, healthy brain contains some 100 billion cells, with each connected to
thousands of other brain cells. The number of possible states that can be assumed is
astronomical, offering a network-redundancy that virtually eliminates the risk of the whole

15
malfunctioning due to a failure of a few individual parts. In addition, the question of which
neuron will be actually activated depends on the distribution of pertinent information in the
whole network at any given time. The potential for “command” is thus distributed over a large
number of components and its location shifts constantly within the network. It is not
permanently localized.
In this regard, the term heterarchy was introduced to characterize a network-like structure
where no vertical hierarchy of authority is discernable. Instead, processes of decision-making
are distributed and determined by function and relevant knowledge rather than by
precedence.
Ultrastability
Ultrastability is another term due to Ashby and the cybernetic concept of regulation. It relates
to the ability of a system to restore homeostatic equilibrium after unexpected perturbations,
even when a trajectory for doing so has not been built-in and fully specified in advance.
Regulation in the cybernetic model is expressed in the context of a system’s capacity to
maintain equilibrium states in the face of disturbances from an environment with which it
interacts. In this sense, regulation can be regarded as the manifestation of a system’s adaptive
capacity. In the simplest case, specifications of perturbation probabilities are built into a
mechanical protective barrier—for example, a wall, a skeletal structure, or a protective shell.
A more complex, dynamic form of adaptation is manifest in the typical homeostatic
mechanism, in which a fixed decision rule is applied to trigger an appropriate corrective action
when equilibrium is disturbed.
Adaptation by ultrastability relates to the more interesting cases—brain-like systems,
societies, or ecosystems—where a sufficient amount of variety is “built” into a system so that
unpredictable changes in its environment can be matched by internal reconfiguration, even
when a specific decision rule is not already embedded in its structure. The more general rule,
instead, is “keep changing internal configurations,” essentially rewiring, in the search for a
subset that matches new demands as they occur.
Even if it is very high, the internal variety of any specific ultrastable system is finite. An entirely
new context-condition may require, or favor, new options that the system, as it is, cannot
generate. In my own doctoral research, incidentally, I was able to extend the cybernetic
concept of regulation to evolutionary processes, cases in which a system actually transforms
into a new entity with a higher regulation potency. Potential variety is amplified in such cases,
for example, by processes of coalition formation, highlighting the importance of cooperation
in evolution.
Reflexivity
In general, the term reflexivity denotes a process that is directed or turned-in on itself. In
cybernetics, reflexivity refers to a circular feedback loop, whereby an observer affects a system

16
 under observation and is affected by it in return. The underlying point is that, contrary to the
orthodox position that stipulates the separation of acts of observation from the observed
phenomena, the two are interdependent and linked in deep ways. The idea was highlighted
by Heinz von Foerster, who proposed the term “second order cybernetics” to account for the
logic of observing systems. The idea was then taken to a radical extreme by Chilean scientists
Humberto Maturana and Francisco Varela, who suggested that observers do not simply review
an “objective,” external reality, but rather they effectively create the world—in a sense in their
own image—by the act of observation.
Scientist, managers, and other active agents are “participant-observers,” they are actors in the
very situations they attempt to understand and control. Reflexivity processes have, therefore,
far-reaching practical implications. The financier George Soros, for example, used the term
with respect to the behavior of financial markets, where investors’ perceptions influence
market behavior only to be influenced, in turn, by market events. Soros claimed that this
theoretical realization guided his trading and the placement of successful financial positions.
Participant-designers in the social realm—change agents of all types—should always keep an
eye on the implications and potentially unintuitive impacts of reflexivity-related phenomena
to their work.
Synergy, self-organization, variety, redundancy of potential command, ultrastability, and
reflexivity are all related concepts. They are significant because they describe the characteristics
of mechanisms that underlie the behavior of complex systems, shedding new light on the concept
of complexity itself. The essential implications are clear: “living,” self-organizing systems, including
social systems of all types, depend on their internal complexity and inherent redundancy for
resilience and long-term viability. It is this internal complexity which allows for the emergence and
reemergence of different configurations in response to changing events.
Plurality, diversity, openness, and agile, responsive structures are the cornerstones of a healthy,
vibrant society and ought to constitute the features of a new world order. A driving, long-term
objective for design interventions in the social domain should be to focus on setting-up smart,
inclusive systems that can adapt, evolve, learn, and continuously self-organize, in the process of
making the world a better place.

10. System Dynamics as a Tool for Engagement

System dynamics is a powerful methodology for understanding the dynamics of complex systems.
It offers a rigorous tool for engagement with systems, a modelling and simulation technique for
analyzing and testing assumptions about a system’s behavior, and a means of policy and strategy
development for active interventions. It is thus effective in guiding efforts of system analysis, as
well as synthesis and design.
System dynamics was originally developed by Jay Forrester and his group at the Alfred Sloan School
of Management at MIT. It was first applied to industrial and corporate issues, but quickly expanded
to address broader social issues as well. Key milestones included urban, health, then world

17
dynamics, culminating in an impactful Club of Rome report The Limits to Growth. The report
addressed issues that today are at the heart of the world’s sustainability agenda. Modelling of
these issues was further developed, to great effect, by Donella Meadows, one of the report’s
original authors. Over the years, the methodology has been continuously refined by the
contributions of numerous practitioners and researchers, and system dynamics groups are now
established in centers around the world. An active System Dynamics Society supports, promotes
and offers a lively platform for further developments.
The system dynamics methodology enhances understanding of a system’s behavior over time by
delineating interdependencies between components, highlighting cause and effect relationships,
and identifying circular feedback structures that dominate behavior. The procedure involves
defining an appropriate system boundary, defining the system’s key variables, mapping the causal
relationships between variables, simulating behavior over time, and exploring significant “what
if?” questions. Identification of operating feedback loops—positive feedback loops which amplify
a gain, and negative feedback loops which reduce it—help pinpoint leverage points in the system
that are critical to bringing about desired change.
Feedback loops and their combined, interacting operations result in archetypal structures which
affect a similar kind of pattern of behavior: increase, decrease, remain level, or fluctuate over time.
Because of their critical effect in driving a system’s behavior, ignoring or disconnecting important
loops can have a distorting and even seriously adverse effect on the whole system. Such critical
systemic disconnects could be responsible for many of the sustainability-ills that now beset our
planet. Think, for example, about the effects of decoupling forces in the economy from biospheric
processes, upon which the economy ultimately depends. Or more specifically, imagine the effects
of ignoring the link between the increasing use of fossil fuels and the long-term impacts of
accumulating greenhouse gases in the atmosphere.

11. Some Illustrative Cases

The power of using system dynamics modelling lies in the possibility of bringing to light hidden or
counterintuitive aspects that would otherwise be ignored. To illustrate, I shall use a few cases from
my own consulting experience, with assignments involving policy and strategy development in
health care, regional planning, and nuclear waste management. These cases were developed in
environments that were alien, and even antagonistic, to system concepts. The following stories
briefly illustrate the effects of utilizing the system dynamics methodology and are meant to convey
a qualitative impression rather than a detailed account of the steps and techniques of the
modelling process itself.
Health Care
This case involves a leading academic medical center in New York, renowned for its excellent,
advanced services and first-class medical education. At the time, hospitals in the United States
were reimbursed at a fixed level for each patient-day, according to a classification based on
multiple criteria, including the type of services offered, their levels of intensity, and the category

18
of institution, whether a teaching hospital, community hospital, urban or rural hospital, and the
like. Health care costs were escalating nation-wide, and there was a strong pressure by
government and insurers to reduce costs.
Management of the medical center came up with an innovative concept for reorganizing its
services in attempt to reduce the average cost of stay per medical episode. The average length of
stay at the hospital was just over ten days, mostly representing serious medical cases. The hospital
rarely had an empty bed, and at any given time, it could rely on a long, constant queue of patients
awaiting admission.
The idea was to construct a simpler medical facility, more like a hotel than a hospital, to which
patients could be transferred towards the tail-end of their stay. At that stage, patients would be
convalescing, and would not require the same level of intensity of services and nursing care. In
addition, patients could move to the new facility, named appropriately “Co-operative Care,”
together with a care-partner, a paying family member perhaps, thus further reducing the need for
constant nursing supervision. At the new facility, patients and care partners could also receive
instructions about the patient’s condition and adjust to new, routine life-requirements before
being released.
The accompanied economic argument was quite straightforward. Assuming that, on average,
three out of every ten days would be spent at the lower intensity and, therefore, lower cost facility,
an incremental savings could occur. The average cost of a typical hospitalization episode would
thus be reduced. Even better, since the hospital would still be paid at a predetermined, fixed per-
patient-day level, established by its category as an intensive medical center, an actual gain could
be registered. The concept looked great all around. The size of the new facility was defined,
architectural plans were completed, financing was secured, and management was ready to move
forward with all the necessary preparations.
The center boasted a forward-thinking strategic planning department—at the time, an unusual
capacity for institutions of this kind—and I was retained to assist in the planning. Some reflection
and basic system mapping quickly revealed that there was a fundamental flaw in the concept, a
flaw that was not immediately apparent given the simple, “linear” assumptions that were made.
The point is as follows: every time a patient would be transferred to the lower cost facility, an
empty bed would become available at the main hospital. Given the long waiting list, it would be
immediately filled with a new patient. The place handled cases requiring highly intensive care, and
the first few days following admission typically represented above-average costs because of
sophisticated diagnostic procedures and frequent surgical work. Furthermore, the hospital
boasted a very active and expensive open-heart surgery program, which because of its prestige
had a strong claim on available beds. Instead of lowering the average intensity and cost across the
board, the ultimate impact of the new program could be exactly the opposite. If followed as
planned, it would actually accelerate the throughput of patients, increase overall intensity of
service, and drive costs through the roof.

19
 Figure 2: A simplified model highlighting some important cause and effect relationships regarding the policy of
creating a co-operative care unit

At first, management resisted this view. To some, it seemed too “intellectual” and unnecessarily
complicated. Some, however, were sufficiently intrigued to allow for a full-blown development of
a system dynamics model. The model mimicked operations of the hospital, mapping the kind of
loops that affected the flow of patients between the two facilities, and the factors affecting
occupancy balance and intensification of services. A couple of examples of such mapping, at low
resolution, are shown in the figures. The full model was developed in great detail, accounting for
each medical service, the flow of patients throughout, the impact of demands on intensification,
and subsequent demands on resources. It was computerized and allowed for the examination of
conditions under a number of different assumptions.
Management embraced the results. The original concept was modified to use the new facility as
an admission facility as well. More careful controls were introduced for optimizing patient mix,
and the model’s predictions with respect to increased capacity needs in radiology, laboratory
services, and surgical rooms were used to guide further planning. Model results helped calibrate
operations of the new program, which was later hailed as an important innovation in health care.

20
 Figure 3: A model of factors affecting occupancy balance and intensification of services in the hospital

Great Lakes Water Level Fluctuations

This case involved work with the International Joint Commission, a joint American-Canadian
organization whose work is focused on the Great Lakes region. It is a high-level commission, with
American commissioners appointed by the U.S. president and Canadian commissioners by the
Canadian prime minister. The commission was originally established to settle issues of water
disputes between the two countries, and over the years the range of issues that came under its
purview expanded. It dealt with questions of water quality and was instrumental in returning the
lakes to pristine health. The work of the commission is supported by a team of Canadian and
American scientists, and it has been closely involved with the U.S. Army Corps of Engineers, which
in the U.S. is responsible for issues related to waterbodies and waterways.
In the mid-'80s, an unusual increase in precipitation during two consecutive years caused water
levels in the lakes to rise. This coincided with a sequence of violent storm events, resulting in major
destruction along the coastline. Lake Shore Drive in Chicago was flooded, and great damage was
inflicted on waterfront properties in both countries. Governments turned to the commission for
advice about how to handle water-level fluctuations. A major study was launched, dominated by
the Corps of Engineers. The underlying perspective was simple: the lakes were misbehaving and
should therefore be contained. Thus, massive protective walls, new canals, and other engineering
works were viewed as the appropriate solution. Members of the scientific advisory staff were
becoming increasingly uncomfortable with this narrow, single minded approach, and I was invited

21
to help think through an appropriate approach to the water level study. This was a case of large-
scale regional planning, which surely called for a comprehensive system perspective. Despite
strong resistance by the leading engineers, enough support was mounting among members of the
staff and the Canadian co-chair of the study. It led to the formation of a system study group, which
I was asked to lead.
A quick mapping focused on the primary components that drove the whole system, and therefore
would be essential parts of an appropriate policy. Even at an early stage, interactions revealed
situations in which following a strict engineering-only approach would result in unintended
adverse consequences. For example, constructing protective measures to contain fluctuation
would prevent wetland flooding, which would then inhibit all of the important environmental
services that wetlands provide. Also, protecting shoreline with structures following damage
inflicted by storms could create a false sense of security. Shoreline development would then
intensify, which would increase vulnerability to possible impacts of potentially stronger storms in
the future.

Figure 4: This model explains how protective structures can bring about a false sense of security that encourages
shoreline development, which leads to increased future vulnerability

The concept of vulnerability drove the development of the model, which was offered as a tool for
assisting the development of government policy. The model integrated three main clusters of
variables: those related to hydrological aspects of water level fluctuations; ecological variables;
and various aspects of human activity, including governance and other characteristics. For some
typical mapping examples, see the figures below. The model was developed to a level of resolution
sufficient to facilitate a well-informed discussion. It drove the crafting of a hybrid long-term
strategy that included consideration of land use management, insurance policy, appropriate
physical measures, and more. These became central in the commission’s final report to
governments.

22
Figure 5: A systems overview depicting interacting primary components, including governance as a key element
related to human activities near the Great Lakes

Figure 6: Detailed overview emphasizing the central role of vulnerability

23
Nuclear Waste Management
This last example involved issues related to the management of nuclear waste in Canada. Canada
boasts large uranium deposits and it developed an active nuclear industry. Several provinces
generate significant amounts of electricity from nuclear technology, and Canadian reactors are
sold around the world. Over the years, the country accumulated large amounts of nuclear waste
that have been stored in temporary facilities, often near the reactors themselves. The material is
extremely toxic, and the toxicity remains active for tens of thousands of years. As in other
countries that generate energy from nuclear fission, the government was seeking a permanent
solution for handling this waste.
The Nuclear Waste Management Organization was then formed, and it was tasked with evaluating
possible solutions and recommending an approach for addressing the issue of nuclear waste. Once
the recommendation was accepted, the organization would become the implementing body to
handle and manage the waste. Halfway through its work, the organization formed a special task
force that was asked to develop a rigorous evaluation methodology, apply it to available options,
and recommend the one that looked the best. Most ideas at the time involved techno-engineering
solutions, ranging from the straightforward to the wild. I was invited to chair this task force, and
the use of the systems dynamic methodology as a key element in the study was accepted from
the start.
From the outset, the model identified four main clusters of variables, the interactions of which
would ultimately drive a preferred management approach. These four clusters covered the
political and economic landscape, factors related to alternative approaches, variables related to
public acceptability, and those related to potential host communities. These were developed to
higher resolution levels, as proven necessary for synthesizing an optimal strategy for the long-term
handling of nuclear waste. The figures below depict the general logic.

Figure 7: Four main clusters of factors that would inform the implemented management approach

24
 Figure 8: A systems perspective of factors leading to implementation of a management approach

A formal, system-based assessment methodology was then applied to evaluating different

options. The integrated results highlighted the need to follow a hybrid, adaptive management
approach. Ultimately, the approach included extensive public engagement and direct involvement
of host communities and was fully adopted by both industry and government. The Nuclear Waste
Management Organization was expanded, funded, and give the task to manage the process. It is
deeply involved in implementation today.
Note the different ways in which system mapping and modelling were employed. In the health
care example, the methodology was used to model, in considerable details, the actual operations
of the hospital, and compute specific values for various variables under different conditions. In
both the water level study and the nuclear waste management case, system dynamics mapping
was used to establish the overall conceptual context and encourage a comprehensive, systemic
perspective in framing the discussion, leading to richer more effective management strategies.

12. Ethical and Practical Implications of the Systems Approach

The implications of the switch from the mechanistic, reductionist perspective to a system view of
the world are profound. The switch is akin to the difference between walking on a straight, hard,
paved surface, to surfing the waves of a heaving sea. This difference forces acknowledgement of
a variable, dynamic, and interdependent reality, and it implies the need for a deep transformation
in values and attitudes, as well as actions in the world. Recognition of the fundamentally
interdependent nature of reality is key, along with the realization that treating a system as though
it were a simple clockwork mechanism is only likely to lead to unintended, even adverse, results.
In the context of an underlying reality of interconnectedness and interdependence, a narrow,

25
mechanistic view is inadequate. It is strategically defective, morally lacking and, in the long run, it
is bound to prove ineffective. It is not the systems way.
Adopting the systems perspective has both practical and ethical consequences. For example:

• Recognition of the complex, dynamic, forever-evolving nature of reality suggests the need
for becoming more accepting of uncertainty, fostering confidence and a level of comfort
with the unknown, and relinquishing an obsessive impulse to control.

• The inescapable reality of change implies the need for agility and resourceful adjustments
in configurations of continuing response. At the same time, the inexorable fact of
complexity calls for respecting complexity for what it is, avoiding trivializing it with over-
simplifications, and mistrusting the quick fix. The inevitable presence of uncertainty
suggests, in turn, the wisdom of a modicum of detachment and a measure of humor,
especially in the face of trials and adversity.

• The pattern of interactions and causal relationships that are disclosed by competent
systems mapping highlight the structures and mechanisms that drive a system’s behavior,
bringing to light the leverage points that are crucial for initiating effective interventions.
The important insight that it is a system’s inner structure that drives its behavior, opens
the door to proactive design as the means of reshaping and reconfiguring structures, in an
effort to ensure preferred outcomes. Systems mapping makes the usually invisible
apparent, and when carried out as a collaborative effort by key stakeholders, it can greatly
facilitate the synthesis and implementation of initiatives for change.

• Accepting a reality of irreducible networks of interactions and interdependencies puts

emphasis on the significance of otherness, and the necessary balance between individual
components and the whole context of which they are a part. The networks of interactions
and the mutual dependencies that these imply compel an appreciation of inclusive,
nurturing relationships, and highlight the role of collaboration and coalition formation in
evolution. Essential interdependencies also alert us to the indispensable role of empathy,
reciprocity, and compassion in healthy interactions.
Integration, synthesis, tolerance, inclusiveness, creativity, and adaptive, hybrid strategies are the
hallmarks of systems-appropriate behavior. Together, these aspects combine to project a
qualitative stance that is fundamentally distinct from the fragmentation, separation, exclusion,
alienation, and short-sighted strategies that are the typical result of the reductionist, mechanistic
approach.
The qualitative aspects that are implied by taking the systems perspective have far reaching
implications to all spheres of human activity. Although starting to spread more briskly in recent
decades, they are still fragile, and they have yet to impact the centers of mainstream thinking.
They are consistent, however, with deep insights of many wisdom traditions that acknowledge the
plurality in unity, promote inclusiveness, and highlight the fundamental prerogatives of moral

26
obligations. They are echoed in the Jewish concept of “Tikun Olam”, or repairing the world; in the
Taoist precept of harmony and becoming one with the rhythms of the universe; in the Christian
precept that the “meek will inherit the earth”; in the Jains principles of harmlessness and non-
violence; and in Buddhist doctrines that betray deep intuitions about the systems nature of reality.
Think, for example, of the basic Buddhist tenets of non-independent existence, non-permanence,
non-dogma, and non-attachment. Contemplate, as well, the concept of Bodhisattvas presented as
idealized beings who embrace a genuine concern for all, and who embody an authentic, all-
inclusive empathy deep in their soul.
Attitudes, values, world views, and the mindsets that drive behavior are of critical importance.
Invariably, they affect the choices we make and the quality of our actions in the world. They can
manifest in a greedy, egocentric, predatory, and domineering orientation or, conversely, they can
find expression in a nurturing, self-restrained, inclusive, and empathetic disposition that
acknowledges the mystery underlying existence and honors the larger system of which we are a
part.
From the viewpoint of sustainability, systems thinking offers a welcomed hope. Because of the
intrinsically systemic nature of the many sustainability-related issues that confront us today,
mastering systems thinking, along with systems-based design skills, is essential for tackling—
perhaps even eliminating—these vexing challenges. The requirement for thinking in systems may
sound simple, even obvious. Yet all too often, in the conduct of personal, social, and global affairs,
it is largely ignored. Ultimately, embracing and internalizing the inherent systems characteristics
of our home-world may inspire the conscious abandonment of a mindset that separates humans
(we) from nature (it) and sees the biosphere as an object that is there to dominate and exploit.
This is the kind of collective transformation that is required in order to bring about a more whole,
creative, healthy, harmonious, and joyful way of being, and a better world.
That completes Module One of the course. To move to Module Two, please click here
(use this hyperlink), or to return to the Table of Contents, please click here (use this one).

27
Further Readings
• Ashby, W. R., An Introduction to Cybernetics, Methuen – University Paperbacks (1964),
(first published 1956).

• Ashby, W. R., Design for a Brain, Chapman and Hall – Science Paperbacks (1966), (first
published 1952).

• Bateson, G., Steps to an Ecology of Mind, Ballantine Books (1972).

• Beer, S., Cybernetics and Management, English University Press (1971), (first ed. 1959).

• Beer, S., Platform for Change, John Wiley & Sons (1975).

• Ben-Eli, U. M., Amplifying Regulation and Variety Increase in Evolving Systems, Progress in
Cybernetics and Systems Research, Vol. VII, Richler, F. and Hanika, F., Hemisphere
Publishing Corporation, (1979), and Journal of Cybernetics, Vol. 9, pp. 285-296, (1979)

• Bertalanffy, L. von, General System Theory, George Braziller (1969).

• Forrester, J. W., Principles of Systems, Wright-Allen Press (1968).

• Fuller, R. B., Synergetics, Macmillan (1975).

• Hock, D., Birth of the Chaordic Age, Brrett-Koeler Publishers (1999)

• Kuhn, T. S., The Structure of Scientific Revolutions, University of Chicago Press, 2nd ed.
(1970).

• Laszlo, E., The Systems View of the World, Braziller (1972).

• Meadows, D. H., Meadows, D., Randers, J., Beyond the Limits: Confronting Global Collapse,
Envisioning a Sustainable Future, Chelsea Green Publishing Company (1992).

• Meadows, D. H., Thinking in Systems, Chelsea Green Publishing Company (2008).

• McCulloch, W. S., Embodiments of Mind, M.I.T. Press (1970), (1st ed. 1965).

• Pask, G., An Introduction to Cybernetics, Hutchingson & Co (1972), (first published 1961).

• Rapparort, R., Pigs for the Ancestors, Yale University Press (1968).

• Yovits, M. C. and Cameron, S., eds., Self-Organizing Systems, Pergamon Press (1960)., pp.
31-48.

• Wiener, N., Cybernetics, M.I.T. Press (1965), (first published 1948).

28
 Module Two
Introduction to Systems Modelling

WASSILY KANDINSKI [PUBLIC DOMAIN], VIA WIKIMEDIA COMMONS

Module Two Overview ​2

The purpose of this particular module is to give you a brief introduction to system dynamics,
where it came from, why it is used, and what are its limitations. We will then move on to
explore the actual building of system dynamics models in the third module.

By the end of this module, you should understand:

What mental models are

Why we use system dynamics
The context in which system dynamics arose
Some of the typical ways in which system dynamics is used
Some of the benefits of using system dynamics
Some of the central assumptions of system dynamics
Some of the limitations of systems dynamics

This module is split up into five sections, described below:

1. Mental Models

2. An Overview of System Dynamics

3. A Brief History of System Dynamics

4. The Central Assumptions of System Dynamics

5. The Limitations of System Dynamics

Section 2.1: Mental Models

Let’s begin this section with a few questions to consider:

Will the increased construction of homeless shelters solve the problem of homelessness in
Dublin, Ireland?

How might increasing access to electricity in Tanzania affect educational outcomes?

How might increasing rooftop solar installations in the US affect the price of electricity from
the grid?

How might lowering working hours affect the material wealth of a country?

Would universal basic income bring about more or less wellbeing in a nation?

Are global warming and the war in Syria in any way related?

How might the increasing use of biofuels affect global food prices?

Can the cutting of trees in a forest reduce that forest's vulnerability to wildfires?

How does your perception of other people affect your behaviour towards them? And how
does your behaviour towards them affect their perception of you?

Exercise
Before we go further, we ask you to spend some time thinking about your own responses to
these questions (the list of questions can be found again below). We recognise that these
questions are quite complex and that you are unlikely to have an answer ready at hand.
Nonetheless, we ask that you make some attempt at answering them, in whatever way you
see fit.

We strongly encourage you to write each answer down on a separate piece of paper, leaving
one side of the page blank. We will ask you to come back to these answers towards the end
of the course to see if your thinking or the way you approach your thinking has changed in
any way based on what you have learned.

You don't need to answer every question, just the ones that are most interesting to you.
We'd ask that you spend no more than 10 minutes in total answering at least two of these
questions.

You are likely to have to make many assumptions in your answers, some of which could be
supported by research (if you feel inclined!) and some of which can be based only on your
own intuition/common sense. We encourage you to explain your answers in as much detail
as you can. However, keep in mind that we are not asking you to give the “right” answers to
these questions (perhaps there is no such thing), but just to use this exercise as a way of
encountering the challenges of answering complex questions and explaining the logic behind
your answer.

Feel free to make your answers specific to your own context. For example, instead of
thinking of Dublin, Ireland in Question 1, you can think about your own city: i.e., Will
homeless shelters solve the problem of homelessness in my city or hometown?

Mental Models
In writing your answers to these questions, you are likely to be using what systems thinkers
would call your "mental model" about the problem or "system" at hand. In this course we will
use the term mental model as follows:

Mental models are deeply ingrained assumptions, generalizations, or even pictures of how
things work that influence how we understand the world and how we take action. (​ Taken
from Peter Senge's 1990 book, The Fifth Discipline)

For example, in response to the previous question about how lowering working hours might
affect the material wealth of a country, you might have said that lowering hours would reduce
a country's material wealth, and that opinion might be based on a certain mental model, a
certain set of assumptions about how the world works. For example, you might have argued
as follows:

“I think that lowering working hours in a country will result in less material wealth for that
country's population. This is because lowering working hours will mean less time is spent
producing goods and providing services, which results in less income generated per person."

Or you might hold a different mental model about the situation, which would lead to a
different answer. For example, you might have argued as follows:

“I think that lowering working hours will result in greater wealth for the population because
the working population will have more time to rest, which makes them more productive in
their working hours. This increased productivity will more than make up for the reduced
hours such that there will be more goods and services produced and therefore, more wealth
to go around.”

Either of these answers and corresponding mental models might be right (or you might
consider both to be wrong) and either might lead to different actions. For example, a person
with the first mental model might vote against legislation for lowering working hours, whilst a
person with the second mental model might vote for it. In general, our mental models about
how the world works tend to inform not only the opinions we hold, but also the policies we
vote for, the decisions we make, and even the way we behave.
What the previous page intends to show is that while most of us do not always build formal
models of how we think, we are still relying on models all the time. The only difference is that
these models are mental models, something that we keep 'in our head,' so to speak. They
are rarely something that we express to others or to ourselves in significant detail.

For example, when we answer a question such as those presented at the start of this
module, we rarely make all/most of the assumptions behind our arguments crystal clear to
others, and even to ourselves. This is because we often hold in our minds such a complex
web of assumptions that it is hard to explain them all by words alone. In many cases we may
not even be aware of this web of assumptions, because we don't take the time to reflect on
them.

Additionally, for many topics we may not have formed any mental models at all. Some
questions/topics may seem too complex for us to deal with, as there are too many things to
consider all at once. Alternatively, we might naively presume that we know the answer
without thinking about it in the depth required.

Summary
That bring us to the end of section 2.1. We hope that by now you understand what we mean
by the term mental model, and can reflect on many of your own mental models about the
world. In the next section we will introduce the methodology of system dynamics.

If you need to return to the table of contents, please click here. We will include these links at
the end of every section to make course navigation a bit easier.

Section 2.2: System Dynamics

System dynamics is a methodology for building formal models of systems, and there are
many different types of systems that system dynamics deals with: anything from an economy
to an ecosystem to a certain part of a healthcare system. Rather than keeping our mental
models about such systems "in our head", system dynamics offers a way to formally map out
those mental models. In doing so, it also offers an opportunity for us to reflect on our mental
models about certain systems, and to perhaps alter them when we see fit, i.e. when we see
a discrepancy between our previous understanding of how a system works, and how it
appears to work based on our new analysis. This allows us to make better informed
decisions that are (hopefully) based on more accurate models about the world.

In this course, we will be focussing on a specific segment of system dynamics, known as

causal loop diagramming. Causal loop diagramming is essentially the qualitative version of
system dynamics modelling, as opposed to the quantitative aspect that deals with simulation
models, i.e. quantified models that use equations to simulate how a system might behave
over time under a given set of assumptions and conditions.

Causal loop diagrams (CLDs) are not quantified. Instead, they are qualitative diagrams that
map out the cause and effect relationships that we assume give rise to a certain
problem/pattern of behaviour. They help us to analyse why certain problems may be
occurring, and why certain solutions may or may not work.

We will begin demonstrating what CLDs look like in the third module!

Complex Systems
System dynamics is typically only used when one is dealing with a complex system. A
system is a very broadly applicable term that can be defined as a number of parts that are
interacting, often with the aim of achieving a certain goal. For example, the human body is a
system made up of your heart, your lungs, your veins, your brain, and much more. All of
these parts are constantly interacting, and together they provide you with the incredible gift
of life. On a more abstract level, a school is also a system, made up of teachers, buildings,
students, books, etc., all working together towards the purpose of providing education.

A system can be complex when there is a a lot of interaction between its parts. This makes it
hard to predict how affecting one part of the system will impact the rest. Note that systems
can be complex even if there are only a few variables: it is the level of interaction between
variables, rather than the number of variables, which makes a system complex.
Nonetheless, more variables in a system can often lead to more interaction, and thus more
complexity.

Examples of complex systems include ecosystems, economies, planet Earth, social

networks, and much more.
Unintended Consequences

The high degree of interdependency in complex system can make it extremely hard to
predict how affecting one part of the system will affect the rest. Certain actions can ripple
along unacknowledged chains of cause and effect, resulting in consequences that are
neither intended nor desired. This explains why many "solutions" to problems in complex
systems may end up being ineffective or, worse still, damaging to the very problem they are
trying to solve.

For example, in the extremely complex system of the global economy, it is often very difficult
to predict how a certain policy will play out. Subsidizing biofuels may go some way to
alleviating the problem of greenhouse gas emissions, but will the increased demand for
crops also put pressure on arable land and increase the price of food, with potentially
disastrous effects?

Another example is ecosystem change, where it is very hard to tell how affecting one
species in the system will affect the rest of the system, given the high level of
interdependency between species in the same ecosystem. Consider how bees play a crucial
role in pollination, and how pollination plays a crucial role in providing future food for bees
and many other species in the same ecosystem.

As such, we see that when a problem arises in a complex system, and some kind of
intervention seems required, it is extremely important that we try to develop an
understanding of the system before intervening in it. In the same way that an investor will
perform “due diligence” on a market before investing in it, we must perform our own due
diligence on a system before intervening in it. System dynamics helps us to perform this due
diligence, so that we find better solutions to problems in complex systems, and avoid - or at
least become aware of - any unintended consequences of our strategies.
Delays, Feedbacks, and Nonlinearities

Apart from a high level of interdependence, systems can also be complex due to the
existence of delays, feedback loops, and nonlinearities. Here we will discuss each of these
system characteristics, and why they make a system complex.

Delays: When there is a significant delay between a cause and an effect, it can be hard to
see the relationship between the two, and to predict how affecting one will eventually affect
the other. Consider how a firm might make orders for new inventory based on existing stock,
but that those orders won't arrive for some time (there is a delay between making the order
and the order arriving), by which time the inventory may be smaller or larger than expected.
This can make the managing of stock a complex affair.

Feedback Loops: feedback loops refer to circular chains of cause and effect. For example,
when an increase in population leads to an increase in births, which leads to a further
increase in population (this is known as a reinforcing feedback loop, a topic discussed in
more detail in the next module).

Nonlinearities: these are relationships whereby an increase in the cause variable does not
produce a consistent change in the effect variable. As an example, consider how an
increase in the price of a good from $1 to $2 might have a much bigger effect on demand
than an increase from $3 to $4.

System dynamics has ways for explicitly highlighting the existence of delays, feedback
loops, and nonlinearities in a system, and this helps one to recognise the complexity of that
system, and find solutions that account for this complexity. Causal loop diagramming is
particularly useful for highlighting feedback loops in a system, and this is something we will
be dealing with extensively in this course.

Group Model Building

Causal loop diagramming is used not only to model your own thoughts, but also the thoughts
of others and of groups that we are a part of. In what is known as Group Model Building,
different stakeholders to a problem gather together and, in a facilitated process, create their
own shared system dynamics model (which can be qualitative or quantitative, or both). By
doing so, a shared understanding of a problem can be developed, power can be distributed
between different stakeholders, and ownership of the solutions identified can be created.
This method serves as a great opportunity for increasing stakeholder participation in
decision making, something which is becoming increasingly recognised as important,
particularly within the sustainability realm.

We will see examples of this method being used later in the course!
The Benefits of System Dynamics

To summarise, below you can see a list of some of the benefits that can arise from
developing a system dynamics model. The model and the process of building it can help us
to:

Think deeply and logically about complex questions/problems, such as those posed at the
beginning of this module
Explain our reasoning/our mental models to others

Make the assumptions in our thinking more explicit to others and even to ourselves

Develop more accurate mental models, i.e. ones that better reflect reality

Forecast different scenarios and identify leverage points that can lead to effective and
long-term solutions

Develop surprising and often counter-intuitive insights about problems, which leads to
innovative solutions

Think about potential unintended consequences of proposed solutions

Sections Summary

That brings us to the end of section 2. We hope that by now you are aware of what system
dynamics is, as well as why and in what context it is used.

In the next section we will take a few moments to understand the context in which system
dynamics arose. This will also give you a feel for the various ways in which it is applied.
Section 2.3: A Brief History of System Dynamics

System dynamics was developed in the 1950s by a man called Jay Forrester. Forrester
originally had a background in engineering and was one of the early innovators in computer
science. Despite his success in these fields, he decided to switch his focus to management
as, in true systems fashion, he felt that what he’d learned in engineering systems could be
also useful in human and social systems. He also felt that management was the realm where
the most pressing problems of the time lay.

Forrester had the opportunity to explore his potential contributions to the field of
management when he was offered a professorship in the newly-formed MIT Sloan School of
Management. His initial goal was to determine how his background in science and
engineering could be brought to bear, in some useful way, on the core issues that determine
the success or failure of corporations.

Jay W. Forrester, August 22 1964; image courtesy of MIT Museum

Industrial Dynamics

Forrester's first experience of the value of systems modelling came from conversations he
had with managers at General Electric. These managers were confused about why the
employment levels in their household appliance plants exhibited three year cycles, whereby
they would have to lay off a significant number of staff every three years. It was easy to say
that business cycles caused fluctuating demand, but that did not seem like a sufficient
explanation to Forrester.

He began to ask about how the corporation made hiring and inventory decisions, and using
a pen and paper he started to draw out the cause and effect relationships that represented
the organisation's hiring policies, and how they related to inventories and orders. From this
model, he could 'simulate' (by doing his own equations on pen and paper) how many people
would be hired on a week by week basis. This first pencil and paper inventory control system
was the beginning of system dynamics. It was described in Forrester's first publication to use
the systems dynamics methodology, which was titled Industrial Dynamics.

Forrester's analysis revealed that even if the firms incoming orders remained constant (i.e.
even if there was no business cycle), employment instability could still arise as a
consequence of common decision-making policies. This highlighted system dynamics' ability
to reveal unintended consequences of policies in complex systems, a theme we discussed a
few pages ago.
Urban Dynamics

In the years that followed, Forrester and a team of graduate students moved the emerging
field of system dynamics forward at a rapid pace, and specific software was developed for
the method, allowing models to be simulated more easily than on pen and paper as
Forrester had originally done.

From the late 1950s to the late 1960s, system dynamics was applied almost exclusively to
corporate/managerial problems. In 1968, however, Forrester began conversing with John
Collins, the former mayor of Boston, who happened to be in the office next to him. From
these conversations, Forrester developed the work known as Urban Dynamics, which was
the first non-corporate application of the system dynamics methodology.

The work was, and is, very controversial, because it shows why many well-known urban
policies are either ineffective or make urban problems worse. Further, the model shows that
counter-intuitive policies - i.e., policies that appear at first glance to be incorrect, often yield
surprisingly effective results. As an example, in the Urban Dynamics model, a policy of
building low income housing creates a poverty trap because such housing used up space
where jobs could have been created, while drawing in people who need jobs. Forrester's
analysis, aided by his model, said that building low-cost housing was a powerful process for
producing poverty, not alleviating it.

This case again demonstrated how system dynamics could be used to discover the
unintended consequences of certain actions or policies, and it also showed how these
models can help generate counterintuitive insights that lead to the discovery of new and
unexpected solutions to big problems.

World Dynamics and the Limits to Growth

Shortly after Urban Dynamics, Forrester was asked by the Club of Rome to use his new
methodology to aid a study on what the Club called the "predicament of mankind": the global
crisis that may appear some time in the future, due to the demands being placed on Earth's
carrying capacity (i.e. its sources of renewable and nonrenewable resources and its sinks for
the disposal of pollutants) by the world's exponentially growing population.

In response to this request, Forrester created a system dynamics model of the world's
socioeconomic system, naming it WORLD1, which was later refined to WORLD2. The model
mapped important interrelationships between world population, industrial production,
pollution, resources, and food. It showed a collapse of the world socioeconomic system
sometime during the twenty-first century, if steps were not taken to lessen the demands on
Earth's carrying capacity. The model was also used to identify policy changes capable of
moving the global system to a fairly high-quality state that is sustainable far into the future.

A further iteration of the model, known as WORLD3, was the basis of the famous book, The
Limits to Growth, in which some of Forrester's students delivered his insights in a more
reader friendly way. Both World Dynamics and Limits to Growth received strong public
reactions, and the discussions they sparked were an important step in increasing awareness
about the necessity of addressing the sustainability challenges facing the planet. The link
between system dynamics and sustainability has been strong ever since.

System Dynamics Today

Before he passed away in 2016, at the age of 98, Forrester spent his remaining years
working on two main things: building a system dynamics model of the US economy, and
spreading the skill of systems thinking and systems modelling in education, from
kindergarten right through to university. Alongside him, countless practitioners have taken
the field further and applied system dynamics in many different industries, including business
strategy, public policy, finance, healthcare reform, environmental management, and most
recently it has even been applied in the modelling of diseases within the human body. Many
of the results have been truly fruitful,

Section Summary

Having given some context to the development of system dynamics, it will be worthwhile to
spend some time thinking about some of its foundational assumptions, many of which are
common to the field of systems thinking in general. This will be the focus of the next section.

Section 2.4: Central Assumptions of System Dynamics

In the pages that follow, we will discuss two of the core tenets upon which system dynamics
is based. These are that structure drives behaviour and that it is better to adapt the
endogenous perspective when solving problems.

Structure Drives Behaviour

We will first discuss the central belief that a system's structure is what drives its behaviour.
To better understand what this means, we can use the iceberg analogy. Systems thinkers
use the iceberg analogy to show that we can look at issues at four different levels: the event
level, the level of patterns, the level of system structures, and at the level of mental models.
The Level of Events
Seeing the world as a series of events is like seeing the tip of the iceberg: you are only
seeing the surface.

For example, imagine that you get a cold. You might react to this event by taking some
medicine to treat the cold. Reacting to events is often necessary, but in the long run it may
be necessary to look deeper to understand why these events are happening in the first
place.

The Level of Patterns

That brings us to the level of patterns, which requires us to look a little below the surface of
everyday events. For example, we might recognise that we are getting a cold every few
months, usually during times when we aren't sleeping as much.
Seeing patterns can help us to anticipate events, which is slightly more effective than just
reacting to them.

The Level of System Structure

If we look deeper still, we will find what systems thinkers call the system structure. This is the
set of cause and effect relationships that produce a certain pattern of behaviour. This does
not only relate to physical structures (e.g. the human body) but also to structures in
organisations (e.g. hierarchies), policies (e.g. a country's constitution), and even cultures
(e.g. the relative importance that a society places on work over leisure).

In the example of getting a cold every few months, you might say that the system that is
responsible for this is your lifestyle. Your lifestyle is essentially determined by the
relationships between things like your work, your leisure, your income, your diet, and your
health. Different relationships between these entities will produce different patterns: a
lifestyle in which you spend more time at leisure and more income on healthy food will likely
result in less colds than a lifestyle that spends more time at work and less income on
improving/maintaining your health.

In system dynamics, the goal of modelling is to reveal the system structure as it is, so that
we can understand why certain patterns and events are occurring.
The Level of Mental Models
If we go deeper still we will arrive at the ultimate cause of the system structure and the
patterns and events we see: this is the level of mental models. As stated previously, mental
models are deeply ingrained assumptions, generalizations, or even pictures of how things
work that influence how we understand the world and how we take action.

In the lifestyle example, we can presume that one will value happiness, and that they will
choose the lifestyle that they think makes them happiest. However, people may hold
different mental models about what makes them happy. If one holds a mental model that
sees their career as the most important thing for their happiness, that sees healthy food as
not worth the money, and leisure not worth the time, then this mindset will ultimately result in
a lifestyle that promotes career over health and leisure. If, on the other hand, we have a
mental model in which we believe health and leisure are as important to happiness as one's
career, then we will lead a totally different and more health conscious lifestyle. This will
ultimately lead to different patterns and events in our life - for example, getting less colds!

The Iceberg model can serve as a useful way of looking at the world. It reminds us to look
beyond the superficial level of events to see the patterns, system structures, and mental
models that have ultimately caused those events. This gives us a deeper level for
addressing issues.
As said previously, reacting to events is like treating the symptoms of an illness without
trying to treat its causes. Recognising patterns can help us anticipate events, but this in itself
does not give us power to control those patterns or events. Changing the patterns requires
us to work on the level of system structure, which in itself is something that almost always
requires us to work on the level of mental models.

For a more enlightening example than catching a cold, consider the collapse of the Lehman
Brothers bank in 2008. This event was part of a larger pattern whereby financial institutions
who were heavily invested in the mortgage industry at the time suffered huge losses, having
previously made huge profits. Systems thinkers would look deeper here and ask about the
system structure that caused this pattern of growth and collapse. They would ask about the
configuration of relationships and incentive structures between regulators, banks,
homeowners and other actors, as well as the values held by actors in the system.

In order to prevent future severe recessions from occurring, it is clear that we would need to
re-design the structure of the mortgage industry, and perhaps of many parts of the financial
system. This in itself will require a significant shift in the mental models of not only policy
makers and regulators, but also the everyday citizen. According to systems thinking, acting
on this level of the system is what will produce the most successful and long term change.

As such we can see that it is our mental models that ultimately determine the way a system
operates. The most effective way to change a system, then, is to change the mental models
of the actors within it. This is really the ultimate goal of system dynamics. Rather than simply
revealing a systems structure, a system dynamics analysis should be considered successful
when it helps people adopt more accurate or useful mental models about the world.

This often requires us to first model the way that the actors in a system perceive that system.
Once these mental models are revealed, we can ask ourselves whether our perceptions
match the reality. If they do not, then we are offered an opportunity to change our perception
(i.e. our mental model) so that it better reflects reality. This in turn leads to better decision
making that is based on a more accurate picture of the world.

Many people would argue that a linear, reductionist, "too-narrow" view of our world still
dominates society, whereby we do not account for the wider system within which many
problems are embedded. In complex systems such as social systems, this often leads to
policies and actions that worsen the very problems they are designed to resolve. System
dynamics offers us a methodology for recognising the actual complexity of real world
systems, rather than shying away from it. This can lead us to discover better and more long
term solutions to problems, and to better communication of the logic of those solutions.
The Endogenous Point of View

We can now move to the second tenet of system dynamics, which is very much related to
the tenet that structure drives behaviour. The endogenous point of view relates to the
boundary that we set when analysing problems (whether through building models or another
method of analysis). It asks us to include within our boundary of analysis all relevant
variables that contribute to the problem at hand, rather than simply saying that the problem
is occurring due to external (exogenous) factors that are outside of our control.

Some examples might make this clearer. In Industrial Dynamics, Forrester showed how
General Electric's internal hiring policies were the cause of the three year cycle in
employment levels, rather than something outside the business, such as fluctuating demand
in the economy. In this sense, Forrester showed that the problem was being generated
endogenously (within General Electric), rather than exogenously (outside General Electric).
By creating a model which purposely set the demand in the economy as constant, Forrester
was demonstrating how the problem (fluctuating need for staff) was not caused by the
economy, but by the managers' hiring policies.

Finding that the source of a problem is within a system rather than external to it is good
news. This is because if the source of the problem is within the system that we can control,
then it is within our power to change it. In the case of the General Electric managers, they
were able to identify how their own policies caused the three year cycle, and so they could
alter their policies to avoid this. If the problem was thought to be external to the system (i.e.
the cycles were caused by fluctuating demand in the economy), then General Electric would
have had less agency to alter the employment cycles.

Implications of the Endogenous Point of View

Adopting the endogenous perspective has some important implications. Many of the big
issues today can only be tackled when looked at from this perspective. Take climate change
as an example. The exogenous perspective on this issue is that the Earth's rising
temperatures are the result of natural cycles outside of human control (such as variations in
solar radiation), meaning that little can be done to change this. The endogenous perspective,
on the other hand, would recognise the role that humans have been shown to play in this
issue (via the burning of fossil fuels, etc.) and would therefore advocate for certain policies to
prevent further climate change, such as subsidies for renewable energy.

We also saw an example of the endogenous perspective in the Great Lakes case, which
showed how the policies for constructing protective structures to counteract the adverse
effects of flooding could actually lead to more housing being developed on that land, and
therefore more vulnerability to such adverse effects. This adopts an endogenous perspective
on a situation where at least one major factor, the water fluctuations of the lakes, was
outside of the community's control.
An exogenous perspective in this scenario would see the flooding as something that simply
happens, and that when it does, one must simply recover and rebuild. It does not recognise
the part that human policies and actions play in the process.

A table that summarizes the differences in perspectives and corresponding policy

implications can be seen below. This was taken from George Richardson's 2011 paper
"Reflections on the foundations of system dynamics", which we recommend as a source of
further reading on this topic.

Sections Summary
Having discussed some of the central tenets of system dynamics, we can now begin to
consider some of its limitations. This is the focus of the next section.

Section 2.5: The Limitations of System Dynamics

In this section we will look at what it means to model a system, and this will lead us to some
of the limitations of modelling in general. We will also reflect on the particular limitations of
qualitative system dynamics modelling in comparison to quantitative system dynamics
modelling.

Model Definition
System dynamics is a technique that uses models to increase understanding of a system.
We have been using the word "model" quite often in this course so far, and so it may be
worth taking a moment to define it. Our definition of a model is as follows:

A model is an abstraction (or a human construct or representation) of reality to achieve a

certain objective.
Aside from mental models, there are many other types of models, used in our daily lives to
make sense of the world. Some of us have used solar system models in elementary school
to understand how planets align around the sun. Some have used maps to navigate through
a new city. Others might have simply drawn human stick figures to communicate a message.

The Limitations of Models

It is important to understand that models are constructed to represent reality, and are not
reality itself. As is often said, “the map is not the territory” (if it was, it would not be a map but
just a big piece of paper that covers everything!). To be useful, models always include some
sort of simplification of the real world. This brings us to the conclusion that:

No model is a perfect representation of a reality

In addition to being simplifications of reality, models are also the product of an observer with
subjective understandings of the world. This leaves them open to biases. Nonetheless,
although models never reflect reality 100%, good models can still be extremely useful in
achieving the objective they were designed for. A map is not the territory, but a good map
still helps us to arrive at our destination. Newton's theory of gravity is not completely
accurate, but it still helps us to predict the motion of large bodies with incredible accuracy.

Similarly, CLDs cannot describe every cause and effect relationship in a system; if they did
they would be too complex to understand, and would not achieve their objective of
simplifying complexity and increasing understanding or communicating insights. However if
the right level of abstraction is chosen, then the model can achieve its objective, whether
that be communicative, analytical, or both. This leads us to the more nuanced conclusion
that:

No model is a perfect representation of a reality, but good models can help us to better
understand and better navigate that reality

The Limitations of CLDs

As mentioned in the Introduction, we will only be focusing on qualitative systems models (i.e.
CLDs) in this course. We will not be focusing on quantitative (i.e. simulation) models. By
simulation models, we mean models that assign values to variables in a model, and use
mathematical equations to describe the relationships between related variables. Using
specific software, one can produce graphs of how the different variables should co-develop
over time, based on the assumptions in your model.

CLDs have advantages in that they are easier to construct and can be more easily
understood in comparison to simulation models. In many cases, building a CLD is sufficient
for the task at hand. However, in many cases it is often necessary to build simulation models
to develop a sufficient understanding of a system. There are several reasons for this, and it
is good to be aware of them. We will discuss each disadvantage in the pages that follow.

Understanding Dynamic Complexity

One disadvantage of CLDs when compared to simulation models is that CLDs are not as
helpful in understanding how variables might interact overtime. When there are several
variables in a model, our minds simply do not have the cognitive capacity to understand how
those variables might co-develop over time, given the numerous relationships between
them, many of which are non-linear. Simulation software does these calculations for us, in
the same way that a calculator helps us do sums that are too challenging to do with our
minds alone.

Nonetheless, we should always remember that simulation models are also limited - they do
not simulate the future, as they rely entirely on the assumptions we put into the model.

Model Validation
Another advantage of simulation models over CLDs is that simulation models allow us to test
our model against reality. If we fill our model with numeric data and then simulate it, we can
see if the simulations of the model match the actual historic data. For example, if we build a
simulation model of the population growth of deer on an island from 1990 up to 2030, we can
see if the model's simulation of the deer population matches the actual historic data (i.e. the
data from 1990 until present), which gives us a good indication about whether or not the
model can be used as a good tool for simulating future scenarios of deer population
development on the island. Given that CLDs do not use numeric data, this means they
cannot be tested in this way, which means there is less opportunity to test their validity.

End of Section 2.5: The Limitations of System Dynamics

Despite these limitations, CLDs can still be very useful tools in many contexts, by helping
individuals and groups to better analyse problems and communicate their understandings
and assumptions. They also serve as a good first step towards building a simulation model,
and can help one to determine whether or not it is necessary to build a simulation model at
all.

Review of Module Two

That brings us to the end of Module Two. To recap, we began the chapter with a series of
complex questions, and asked you to spend some time thinking about some of these and
writing down your answers. We then explained how systems thinkers would say that these
answers are based on your mental model about those topics/problems. From there we
discussed what the term mental models means, and talked about how it can be difficult to
become aware of or change our own mental models about complex topics. We also
discussed how communicating our mental model to others can be very challenging.

System dynamics was then proposed as a tool to help us formally model ours and others'
mental models. It was said that system dynamics models can help us to both develop and
communicate our mental models about complex systems. In other words, they help us to
develop and communicate our understanding of how the world works.

Following a brief overview of the origin and history of system dynamics, we touched upon
two of its central tenets: that structure drives behaviour, and that it is better to adopt the
endogenous point of view when addressing problems. After reviewing the benefits of system
dynamics modelling, we noted some of its limitations, and particularly the limitations of
qualitative as opposed to quantitative modelling.

We hope that by now you are excited to learn about system dynamics. In the next modules
we will begin demonstrating what CLDs look like, how to read them, and how to build them.
More importantly, we will demonstrate how building these models can help us to better
understand complex problems, and better identify the most effective and long-term solutions
to our and our society's challenges.

System dynamics is essentially just a technique. However, the use of this technique can
radically transform the way in which one approaches problem, and even the way in which
one understands the world. We hope that this course will not only offer you a new technique,
but also offer you an opportunity to look at the world in a new and hopefully clearer light.
 Module Three
The Basics of Systems Modelling

WASSILY KANDINSKY - CIRCLES IN A CIRCLE, 1923

Module Three Overview ​3
In the previous module, you have learned that we build CLDs to help us understand and find
solutions for issues in complex systems, as well as to communicate our assumptions more
explicitly and show the logic behind our thinking and/or any solutions we propose.

In this module, we will focus on the basic building blocks of CLDs. Explaining how they use
words and arrows to display the different cause and effect relationships and feedback loops
that are assumed to give rise to a certain problem or pattern of behaviour.

Module Outline
The module is divided into four sections, as described below:

1. Casual Relationships in CLDs

2. Feedback Loops

3 Behaviour Over Time Graphs

4.Systems Archetypes

Module Goals

By the end of this module you should be able to:

Read causal relationships in CLDs.

Understand and identify the different kinds of feedback loops (i.e., reinforcing and balancing)
in a model

Read behaviour over time graphs

Identify and describe systems archetypes and leverage points

Section 3.1: Causal Relationships in CLDs

If you consider casual loop diagramming to be a language, then you can think of the next
section as teaching the basic grammar of this language. And while grammar is not always
everyone's cup of tea, it is a necessary step on the way to speaking and understanding a
language fluently. So please bear with us if the next few pages aren't so captivating; it will be
worth it eventually!

Two Types of Causal Relationships

A cause and effect relationship in a CLD is represented by an arrow between two variables.
By “variable,” we mean literally anything that can change over time. The arrow goes from the
cause variable (the one causing the change) to the effect variable (the one being affected by
the change).

There are two types of cause and effect relationships in CLDs: positive and negative. A
positive relationship is depicted by the '+' symbol at the arrow's head, whereas a negative
relationship is depicted by the '-' symbol at the arrow's head. In this course we will use also
use blue arrows for positive relationships and red for negative ones, although this is done by
all system dynamicists.

Note: Remember, to zoom in and out of model pages, use the buttons in the top right corner
of the page. The symbol with two arrows will fit the model zoom to capture the whole model.
A positive relationship means that the cause and effect variable are moving in the same
direction, such that an increase in the cause variable results in an increase in the effect
variable, and a decrease in the cause variable results in a decrease in the effect variable, all
else equal.

Note that the term positive does not mean that the relationship is necessarily a desirable
one. Instead the term positive here comes from its mathematical use, and only means that
the two variables move in the same direction.

Conversely, in a negative relationship, the effect variable moves in the opposite direction to
the cause variable. So in this case an increase in the cause results in a decrease in the
effect, and a decrease in the cause results in an increase in the effect, all else equal.

Again, the term negative does not mean undesirable but rather refers only to the fact that the
two variables move in the opposite direction.

Some examples will help make things clearer:

Example of a Positive Relationship

You may remember the two mental models we expressed in Module Two relating to the
question of how lowering working hours might affect a country's material wealth. In the first
mental model, the opinion was expressed that lowering working hours will result in less
material wealth because there will be less goods produced and less services provided.

The causal relationship expressed in words here can be formally modelled as seen below.
This causal relationship is saying that if the Hours worked per week were to be decreased,
then there would be less Goods produced/services provided per week. It is also saying that if
the Hours worked per week were to be increased, then the Goods produced/services
provided per week would also increase, all else equal.

Another example of a positive relationship is Births to Population. This is a positive

relationship because an increase in Births will mean that the Population will be larger than it
would be if the Births hadn't increased. Additionally, a decrease in Births will mean that the
Population will be smaller than what it would otherwise have been.
... what it would otherwise have been
On the previous page we used the phrase “...what it would otherwise have been". This
phrase is applicable to all causal relationships in CLDs. We say this because a decrease in
the cause variable does not necessarily mean a net decrease in the effect variable (in a
positive relationship). It just means that the effect variable will be less than it would otherwise
have been (i.e. if the cause variable hadn't decreased). Additionally, in a positive relationship
an increase in the cause variable does not necessarily mean that the effect variable will
increase, it just means that the effect variable will be more than it would otherwise have
been (i.e. if the cause variable hadn't increased).

For example, if the births in an area decrease from 10,000 a year to 7,000, the native
population will still continue to grow as long as births remain greater than deaths in the area.
However the decrease in births means that the population will grow at a slower rate. As
such, the decrease in births (the decrease in the cause variable) means that the population
(the effect variable) will be lower than it would otherwise have been.

All Else Equal

You may have also heard us use the term all else equal when describing the meaning of
causal relationships. This is because causal relationships in CLDs are always considered in
isolation to the rest of the system. When we determine each causal relationship, we always
assume that all other variables in the system will remain equal. You may hear some systems
thinkers use the term "ceteris paribus", which is Latin for "all other things being equal".

So based on the model below, we can say that more Access to healthcare will lead to a
greater Life expectancy, ceteris paribus.

Of course, Access to healthcare is not the only variable that affects Life expectancy. For
example, one's Quality of diet will also affect their Life expectancy. This can be modelled as
seen below:
As such, when reading or developing causal connections in CLDs, we always
determine/read the type of relationship in isolation to the other variables (i.e., assuming that
they remain equal). So the above model is saying that you an increase in Access to
healthcare will increase your Life expectancy, without regard to what is happening to your
Quality of diet, which is a separate variable and so modelled separately.

Exercise 3A: Positive Relationships

Can you think of a positive relationship? It can be related to your personal life, to your own
studies or profession, or to something that occurs in nature.

The important thing to keep in mind is that an increase in the cause variable will result in the
effect variable being more than what it otherwise would have been, and vice versa.

Some examples of positive relationships include those summarised in the table below.
Example of a Negative Relationship

Below, you can see an example of a negative relationship. Here we are saying that if the
Price of grapes increases, then Demand for grapes will decrease (due to some assumed
price sensitivity of people who buy grapes). We are also saying that if the Price of grapes
decreases, then the Demand for grapes will increase.
Another example of a negative relationship is that of Deaths to Population. An increase in
the number of Deaths means that the Population will be smaller than what it otherwise would
have been (i.e., if the Deaths hadn't increased).

If Deaths decrease, then the Population will be larger than what it otherwise would have
been.

Exercise 3B: Negative relationships

Can you think of a negative relationship?

It can be related to something that occurs in nature, to your personal life or to your own
studies or profession.

The important thing to keep in mind is that an increase in the cause variable will result in the
effect variable being less than what it otherwise would have been, and vice versa.
Answers to 3B
Some examples of negative relationships include those summarised in the table below.
Causality not Correlation

Another important thing to note is that in CLDs, we portray relationships that we assume are
causal, rather than those that are only correlated. An example might help us to illustrate the
difference between correlation and causality: a famous study found that there is a strong
correlation between ice cream sales and the murder rate on a given day. However there is
no sensible way that we can imagine ice cream sales influencing the murder rate, or the
murder rate influencing the sales of ice cream! A more sensible conclusion is that something
else is at play.

Indeed, ice cream sales are likely to increase on a hot day. At the same time, studies have
suggested that hot weather increases the murder rate on any given day, for a number of
reasons. For one, more people are out of their homes, which increases the chance of conflict
(for more on this theory, see this article). As such, it seems that temperature is the real
causal variable that affects both ice cream sales and the murder rate.

Incorrect model:

Correct model:
The important insight from this is that when creating CLDs, it is important that we only
portray what we assume are causal relationships, rather than relationships of correlation.

Exercise 3C: Causal Relationships

Bellow you can see three variables (Interest payments, Withdrawals and Money in bank
account).

Can you draw the arrows between the variables to depict the correct causal influences? And
can you determine whether each relationship is positive or negative?

The answers are on the next page.

Answer to 3C

An increase in Interest earned (i.e. interest that you receive from the bank) leads to the
Money in bank account increasing above what it would otherwise have been. Similarly, a
decrease in Interest earned lead to less Money in bank account, compared to what would
otherwise be the case. As such, Interest earned to Money in bank account is a positive
relationship.

An increase in Withdrawals leads to less Money in bank account, and a decrease in

Withdrawals leads to more Money in bank account, (compared to a case where Withdrawals
hadn't decreased). As such, Withdrawals to Money in bank account has a negative
relationship.

You may also have drawn a positive relationship from Money in bank account to Withdrawls,
under the assumption that the owner of the bank account will take out more money when it is
available. We will discuss such relationships later in the module!
A Note on Naming Variables
When you start building your own CLDs, it is important to remember some best practices
that will make your CLDs as useful as possible. Two of the best practices in naming your
variables are:

Variables should be nouns or noun-phrases rather than verbs.

Choose variable names whose normal sense of direction is positive.
Let's explain why these are considered best practice.

1. Variables should be nouns or noun-phrases rather than verbs.

This is because the actions (verbs) are captured by the causal links connecting the
variables. So for example, it is better that we have variables such as Costs and Price rather
than variables with names like Costs rise and Price rises. Adding the verb rises to the
diagram presumes that costs will only rise, which biases our discussion towards one pattern
of behaviour (increasing costs and prices).

Not best practice:

Best practice:
2. Choose variable names whose normal sense of direction is positive
So for example, it is better to have a variable called Happiness than to have a variable called
Unhappiness. This is because it is easier to think of an increase/decrease in Happiness
rather than an increase/decrease in Unhappiness. And it is generally better to have a
variable called Profits rather than Losses, because it seems clearer to talk about falling
profits rather than falling losses.

Not best practice:

Best practice:

There may be some exceptions to this practice, but as one of the most prominent system
dynamicists, John Sterman (of MIT), once said: "decreasing noncompliance with this
principle will diminish your audience's incomprehension"!

Exercise 3D: Errors in Variable Names

Try to spot the errors in the variable names from the examples below:
Answers to 3D
Not best practice:

Best practice:

Not best practice:

Best practice:
End of Section 3.1: Causal Relationships in CLDs

That concludes the first section of this module. We hope that by now you:

Understand the meaning of positive and negative causal relationships

Understand that each causal relationship in a CLD is determined in isolation to the rest of
the system (i.e., all else is considered equal)
See that relationships in CLDs are meant to signify causality between two variables, rather
than correlation.
Are aware of best practices for naming variables
In the next section, we will begin looking at feedback loops.

Section 3.2: Feedback Loops

Now that we've gotten through the basic grammar of how to read and write causal
connections in CLDs, we can finally move on to demonstrating how CLDs can help us to
analyse a problem—in other words, to speak the language!

One of the most valuable things about constructing CLDs is that they help us to recognise
feedback loops in the system at hand.

This section will describe what feedback loops are, and why it is so important to recognise
them in a system.

Linear Chains of Cause and Effect

Up until now we have only modelled one-directional (also referred to as unidirectional)
relationships between a cause variable and an effect variable. Don't worry, things will get
more interesting!

In systems thinking, we of course recognize that effects are often causes to other variables.
For example, more Organic farms leads to less use of Chemical pesticides, which positively
contributes to Biodiversity, which in turn leads to higher Soil quality. Modelling these chains
of causes and effects in CLDs can help us to recognize how certain variables are
interrelated even though the relationship—or at least the mechanism by which that
relationship exists— might not be obvious at first.
This is one of the benefits of causal loop diagramming. It aids our thinking, by having us map
out causal relations that may otherwise be difficult to recognise, or to explain to others.
Circular Chains of Cause and Effect

When we follow a chain of cause and effect, we might often end up back where we started!
Our last variable feeds back to the original cause. This means that we have discovered a
feedback loop in the system.

A feedback loop is a circular chain of cause and effect. So when A affects B, B affects C,
and C affects A again, we say that a feedback loop exists between these variables.

Just as there are only two types of causal relationships (positive or negative), there are also
only two types of feedback loops: reinforcing or balancing.
Reinforcing Feedback Loops (R)

An example of a reinforcing feedback loop is the one of Births and Population.

The positive relationship from Births to Population also works the other way around - the
higher the Population, the more Births there will be, all else equal.

Therefore, an increase in Population will result in an increase in Births, and so a further

increase in Population. The initial increase in Population is reinforced by this feedback loop
structure in the system.
Reinforcing Feedback Loops (R)

Another example of a reinforcing feedback loop is that of the Ice Albedo effect.

An increase in Greenhouse gas (GHG) emissions per year leads to more GHGs
accumulating in the atmosphere and more Heat trapped. Because the atmosphere gets
warmer, more Ice melting will occur. More Ice melting leads to more open water which
increases Sunlight absorption. More Sunlight absorption, in turn, leads to even more Heat
trapped.

As you can see, reinforcing feedback loops are depicted with an R in the middle of the loop.
If it's useful, we can sometimes give our feedback loops names, as we've done here.

Tip: to see all the variables that are part of a feedback loop, simply hover your cursor over
the feedback loop name, and the relevant variables will be shown.
Reinforcing Feedback Loops (R)
To summarize, in a reinforcing feedback loop, a change of direction in one of the variables
leads to even more change in that direction. As such, we can say that reinforcing loops
always cause a greater amount of change in a system—they reinforce change, so to speak.
This is why it is so important that we are always aware of reinforcing feedback loops in any
system. The construction of a CLD is an effective way of identifying feedback loops in a
system.

The change brought about by reinforcing feedback loops can sometimes be desirable, and
sometimes undesirable. When the change is undesired, we say that there is a “vicious loop"
(or vicious cycle) at play. When the change is in a desired direction, we say that there is a
“virtuous loop” (or virtuous cycle) at play.

Exercise 3E: Reinforcing Feedback

Let's try and see if we can spot some feedback loops. Can you draw the correct causal
connections and make a reinforcing loop with the variables shown below?

The answer is on the next slide.

Answer to 3E
The Money in bank account determines the amount of Interest earned: the more Money in
bank account, the higher the Interest earned. Higher Interest earned means more Money in
bank account. As such there is a reinforcing loop between the two variables.

Conversely, if the Money in bank account reduces, then so do the Interest payments, which
means even less Money in bank account, compared to what would otherwise have been.

Note that the Interest rate is considered constant in this example and that there are no
withdrawals. Remember the ceteris paribus concept?
Balancing Feedback Loops (B)
The other type of feedback loop that exists is balancing loops. Whilst reinforcing loops
generally cause greater change within the system, balancing loops generally do the opposite
- they prevent change and create greater stability in the system, whether that be desired or
not.

An example of a balancing feedback loop is that of Deaths and Population. When a

Population increases, the total number of Deaths also increases, all else equal. With an
increase in Deaths, the Population in turn decreases. An increase in Population is thus
balanced by an increase in Deaths. Conversely, a decrease in Population will be partially
compensated for by a decrease in Deaths.
Balancing Feedback Loops and Goals
Whenever there is a Desired state in a system, then there will almost always be a balancing
loop that describes how that goal is achieved.

We say that as the Gap between the Desired state and the Current state widens, there will
be more corrective Action. This brings the Current state closer to the Desired state, hopefully
until the Gap between the two is closed.
Balancing Feedback Loops (B)
For example, when the Temperature in an apartment is lower than the Desired temperature,
then there is a Temperature gap. The heating system will sense this (via a thermostat) and
will generate Heat such that the Temperature gap is eventually reduced to zero (i.e, such
that the Temperature equals the Desired temperature). If the Temperature drops below the
Desired temperature again, then more Heat will be generated until they are again equal. This
is a balancing loop.

Feedback loops (both reinforcing and balancing) can be found in all walks of life, from
psychology to ecology. Before moving on to the next section, we recommend that you watch
the excellent video linked below, which should help you get a better grasp of how important
and how pervasive feedback loops are in our lives and our planet.
End of Section 3.2: Feedback Loops

That concludes Section 3.2. We hope that by now you:

● Understand that there are two different types of feedback loops: reinforcing and
balancing
● Can determine whether a feedback loop is reinforcing or balancing
● Understand that reinforcing loops promote change in a system, whereas balancing
loops negate change/promote stability
● Understand that the way in which goals are achieved can often be represented by
balancing loops
● Begin to recognise feedback loops around you, in nature, society, the economy, and
even in your life in general

In the next section we will look at another tool for helping us understand systems: behaviour
over time graphs.
Section 3.3: Behaviour Over Time Graphs

Now that we have an understanding of what reinforcing and balancing feedback loops are,
its time to start thinking more deeply about what effects they can have on a system. One of
the best ways to do this is to draw or visualise behaviour over time graphs. These are
graphs over time of the most important variable(s) in a model. They demonstrate how the
variable(s) is likely to develop over time, given the system structure that it is part of. If we
refer back to the Iceberg analogy, a behaviour over time graph essentially illustrates the
system on the level of patterns.

Exercise 3F
Lets look again at the system structure between Interest earned and Money in bank account.

What do you think happens to the Money in bank account over time? Assume that it starts
as 1000 USD in year 0, and that the Interest rate is 0.4 and is compounded annually (hint:
this means that the Money in bank account is doubled approximately every 2 years - this is
an unrealistically high interest rate but is used for illustrative purposes!).

How large will the Money in bank account be twenty years from now, assuming it grows
unhindered? Make a quick guess (no need to get out a calculator) and sketch it on a piece of
paper with a graph similar to what is seen below:
Answer to 3F

The graph below shows the behaviour of the Money in bank account over time (don't worry,
we most definitely used a calculator to get this!). In twenty years the Money in bank account
grew from $1000 to almost $900,000, without any deposits from the owner!

This behaviour is called exponential growth. Exponential growth is created by reinforcing

feedback loops as the change of a variable is amplified over time. The more Money in bank
account, the more Interest earned, and the more Interest earned, the more Money in bank
account!

Many people underestimate the power of exponential growth. To see a more visual
illustration of its power, imagine that you were to put one grain of rice on the first square of a
chess board, two on the second, four on the third, eight on the fourth, following this pattern
all the way up to the 64th square on the board. How much rice do you think you would need
to do this?
Other Examples of Exponential Growth

The growth of a bank balance through interest payments is just one example of exponential
growth driven by a reinforcing feedback loop. There are many others. For example, as we
have seen, a population grows via the reinforcing feedback loop between births and the
population (the more births, the more people, the more people, the more births). This goes a
long way to explaining why the world population's historical growth looks the way it does,
exhibiting a clear pattern of exponential growth.

This also shows us how a system's structure is what produces its behaviour. Although the
money in a bank account and the global population seem like unrelated systems, they have
similar structures (i.e. they are dominated by reinforcing feedback loops) which means that
they exhibit similar patterns of behaviour, which in this case is exponential growth.
Goal-seeking Behaviour
The structure of balancing feedback loops produce a pattern of behaviour which we call
"goal-seeking behaviour". To demonstrate what this means, let's look again at the thermostat
example:

Suppose the Temperature in the apartment is 5 °C degrees while the Desired temperature is
25 °C. The Temperate gap between the desired and actual temperature will cause
thermostat to generate Heat such that the Temperature will gradually approach its goal of 25
°C. Initially, the gap between the Temperature and the Desired temperature is large and as
such the heating system will work full power to make up for this gap. However, as the
Temperature gap gets smaller, the heating system will not need to provide as much Heat. As
seen in the behaviour over time graph below, the increase in Temperature is significant at
the beginning (as the heating system works full power) and gradually becomes less and less
as the Temperature reaches the goal of 25°C.
Oscillation

When there is a delayed cause and effect relationship in a balancing loop, this can result in
what is known as oscillating behaviour.

You may have experienced oscillation whilst in the shower. When the water is too cold, you
might open the hot tap more. However this doesn't cause the water to heat up immediately,
as the heating system takes time to react to your adjustment. This might cause you to
impatiently open the hot tap even more, such that when the water does eventually arrive, it is
far too hot! This might then cause you to quickly turn the hot tap way down, with the result
that the water eventually arrives as too cold!

This structure of this system can be represented as seen below. Note that the double
dashed arrow from Shower faucet setting to Water temperature represents the fact that there
is a significant time delay between the adjustment of the faucet setting, and the change in
the water temperature. In other words, the double dash on an arrow means that there is a
significant delay between the cause and effect.
Below we can see a behaviour over time graph that depicts how the temperature of the
water might oscillate around the desired temperature (the dotted line), based on this system
structure.

S-shaped Growth

Many if not most systems are made up of both balancing and reinforcing feedback loops.
The existence of these two kinds of loops in a system can result in what is known as
"S-shaped growth," which can be seen in the graph below.

A variable exhibits S-shaped growth when it is at first dominated by a reinforcing feedback

loop, and then later dominated by a balancing loop as it reaches some kind of limit to growth.
Systems thinkers refer to this situation as a shift in loop dominance in the system.

For example, consider how a population of deer might initially grow exponentially (dominated
by the feedback loop between births and population), and then, as the population grows
towards the carrying capacity of its area, the death rate of the deer might increase (or the
birth rate might decrease), such that the death rate equals the birth rate, and the population
of deer reach what is known as a dynamic equilibrium (i.e., a steady state).
Exercise 3G: Feedback Loops

Take a look at the CLD on the below. It is a generic model of how a firm's Marketing efforts
affect their Sales.

● Q1: Can you put the correct signs next to the arrows?
● Q2: Can you identify and define the type of feedback loops present?
● Bonus Question: Once you have answered Q1 and Q2, can you sketch the behaviour
of the firms Sales on a graph over time? Assume that the Market saturation is at first
low (imagine the firm has just launched a new product), but eventually grows to full
saturation over time.
Answers to 3G (1/3)

In this system there is a reinforcing and a balancing loop. The reinforcing loop says that an
increase in Marketing efforts increases Sales, and as Sales and Revenue continue to grow,
the company has the ability to further increase its Marketing efforts.

If we had a mental model that only recognized this loop, then we would presume that
increasing Marketing efforts always led to increases in Sales.

Tip: to determine if a loop is reinforcing or balancing you can:

1. Follow the loop and the logic of each relationship. For example, start at one variable
and imagine what will happen if you increase that variable. If you follow the loop
around and an original increase in that variable leads to another increase, then the
loop is reinforcing. If it leads to a decrease in the variable, then the loop is balancing;

OR

2. Count the positive and negative signs. A balancing loop always has an uneven
number of negative signs (at least one), whilst a reinforcing loop has either no
negative signs, or an even number of them.
Answers to 3G (2/3)
However when we look at the wider system in which Marketing efforts operate, we see that
increased marketing will only lead to increased Sales for so long, because when Sales
increase, then the level of Market saturation also increases. This is because everyone who
would potentially buy the product (i.e. the addressable market) is already doing so. This
highights a balancing loop in the system.

At some point the market will reach total saturation. Increasing the Marketing efforts then
has less effect on Sales (it does not cause Sales to increase, as before, but rather just keeps
the same amount of people buying the product). This may seem strange to the management
team, as increasing their Marketing efforts previously always led to more Sales.

Of course many companies who have huge sales still engage in marketing. And this model
is not saying that they are foolish to do so. The likes of Apple and Microsoft still need to
advertise their products if they want to keep the customers they have won. The point this
model is trying to make is that marketing executives at growing companies should not be
totally surprised when the effectiveness of their marketing efforts (in terms of generating an
increase in sales) begins to decline as their product reaches market saturation.
Answers to 3G (3/3: Bonus)

id you estimate that Sales would develop in the way shown on the graph below? Our thought
process behind reaching this answer went something like this:

"Initially, there is no market saturation, which means the balancing loop is not so strong. This
means the reinforcing loop is likely to dominate, resulting in an exponential growth in sales.
However, growth in Sales eventually leads to significant Market saturation, such that Sales
slow down and finally stabilizes when full Market saturation (shown by the dotted line on the
graph) is reached (i.e., when the entire addressable market already owns or is consistently
buying the product)."

As such, the system shows S-shaped behaviour, which is typical for systems with a
combination of a reinforcing and a balancing feedback loop.
End of Section 3.3: Behaviour Over Time Graphs

That concludes section 3. We hope that by now you:

● Realise the value of behaviour over time graphs in understanding how a systems
structure leads to its behaviour
● Understand the power of exponential growth, which is typically driven by a reinforcing
feedback loop
● Understand that balancing loops often produce goal seeking behaviour
● Understand that delays in balancing loops can often result in oscillation around a
goal
● Understand that a combination of a reinforcing and a balancing loop often leads to
S-shaped growth
Section 3.4: Systems Archetypes

In this section we will look at what are known as systems archetypes.

Limits to Growth: A Systems Archetype

In the scenario of the previous exercise, more Efforts initially led to improvements in
Performance (measured by sales) (R). However, over time that Performance reached a limit
or Constraint that caused the Performance to slow down (B).

Recognising this scenario leads to the understanding that you should either look for ways to
remove the limits or else prepare to adapt to them. For example, the company in the
previous exercise could have explored new markets to increase its addressable market
(removing the constraints of market saturation), or could have planned (e.g. budgeted) for
the fact that marketing efforts were unlikely to result in the same growth in sales as was
previously experienced.

Archetypes
The scenario in the previous page is an example of a systems archetype (Limits to Growth).
Systems archetypes are combinations of reinforcing and balancing feedback loops that
commonly occur across different systems.

Each archetype has a characteristic theme and storyline, behavioural pattern, and potential
for action. Being able to recognize system archetypes in different situations allows for a
deeper and faster understanding of that system, and could also help us design effective
interventions. By applying archetypes habitually, you will learn to “see” structures when you
hear stories that are similar to the archetype.
Let's look at some more examples.
Tragedy of the Commons Archetype
Imagine a small lake in rural Vietnam where everyone is allowed to fish. The lake doesn’t
belong to anyone in particular and you can fish as much as you want. Imagine that there are
two fishing companies in the area and that all those who fish in the lake belong to either
company A or company B. As can be seen on the right, the more fish company A catches,
the more profits they make and the more they will increase their Fishing (R1), as profits allow
the company to hire more fishermen and buy more fishing boats. This principle is the same
for Company B (R2) and together both companies make up the Total fishing.

In the beginning both Company A and Company B make good profits and provide a valuable
service to the community.
However, when we look at the longer term we see that this system, if left uncontrolled, may
result in some unintended consequences. If the level of fishing is not limited in some way,
then it may eventually lead to a situation whereby more fish are being caught than are being
naturally replenished, leading to a decline in the number of fish in the lake.
Less fish in the lake in turn leads to more effort to catch a fish, as fishermen might need
either better boats to go to deeper waters, or to spend more time fishing to catch the same
amount of fish. This reduces the profits of both companies, as well as the fish supplied to the
community. It also damages the fish population and perhaps the whole aquatic ecosystem.
The scenario in the previous example describes the "Tragedy of the Commons" archetype
and is often referred to when people discuss environmental issues (the phenomenon
became widely known through an article by ecologist Garrett Hardin in 1968). In the Tragedy
of the Commons archetype, individual actors each behave rationally according to their own,
short-term self-interest (R1 and R2) but that eventually results in the depletion or spoiling of
a shared resource (B1 and B2), which is bad news for all involved.

Possible solutions to Tragedy of the Commons type problems are to:

1. Educate actors about the consequences for the population as a whole

2. Reduce the activities by regulating access to the resource or privatizing the resource
so that the owner take longer term considerations into account.
3. Reduce the information delay between the total activity and the resource being
depleted.
The below video gives a more in depth and colourful description of the Tragedy of the
Commons. If you have the time, we suggest that you watch it!

Success to the Successful Archetype

On the right you can see another archetype, known as the "Success to the Successful"
archetype. This archetype explains how one actor in a system can become far more
successful than other actors due to small differences in initial conditions, expectations, or
random factors.

For example, when the coach of a football team thinks that player A is better than player B,
he or she will spend more time training player A. The time of the coach is limited, so player B
will receive less training than A. Because player A receives more training/attention, he or she
might improve even more than player B, further widening the gap in the skills between them.
This will further justify the coach's decision to train player A instead of player B.

This self-fulfilling prophecy continues to reinforce the success of player A while diminishing
the success of player B, all because of an initially small difference in skill between them.
This archetype applies to a variety of situations and can explain for example:

● Why there is social and economic inequality within and between many countries
● Why the rich are getting richer and the poor are getting poorer
● How one species can drive another to extinction
● Why success at school leads to success later in life
There are two main strategies for dealing with similar situations:

1. Redistribute the unequally distributed resource. An example of this is progressive tax

scales or property taxes.
2. Separate the two reinforcing loops so that the two entities are not competing for the
same resource anymore. For example, if a fishing community who was previously in
competition with another began fishing in a new lake.

Exercise 3H: Archetypes

We encourage you to watch the three videos below. Can you identify which systems
archetype is applicable to each of the news stories? The answers are on the next page.

1. Delhi chokes on air pollution

2. End of the outbreak: Animated look at the ebola epidemic

 3. What discrimination looks like in America

Answers to 3H
The "Tragedy of the Commons" archetype is applicable to the first video. In this case, the
shared common resource is clean air. This essential resource is being spoiled by individual
actors that behave rationally according to their own self-interest (by driving cars or burning
crops, for example), despite the fact that spoiling the resource is contrary to each others'
long-term best interest (public health).

The Limits to Growth archetype is applicable to the second video. In this case, the virus
(Ebola) initially spread rapidly but was eventually slowed down when there were less and
less people to infect, as people were either already infected or those infected were put into
quarantine/isolation.

The Success to the Successful archetype is applicable to the third video. Due to initial
prejudices and discrimination, a minority group (black people in the US) became socially
disadvantaged. This social disadvantage was then interpreted not as the result of earlier
prejudice and discrimination but as evidence that the minority is indeed inferior, which
resulted again in prejudice and discrimination.

End of Section 3.4: Systems Archetypes

As you can see, archetypes can be used to identify the systemic structure of a problem and
to remind us about potential high leverage solutions to those problems. When you become
familiar with the different systems archetypes, you might start seeing them everywhere. You
might understand humankind's history through the lens of the Success to the Successful
archetype, or you might see a messy kitchen as the result of a Tragedy of the
Commons-type structure between yourself and your housemates!
Being aware of archetypes can help us to identify the systemic structure of a problem and to
remind us about potential high leverage solutions to those problems. Relating a problem to
an archetype can help you to see how a problem might be similar to many others, and can
help you to see common solutions to those problems. For example, you might recognize that
the collapse of fish stocks and global warming might both be results of a "Tragedy of the
Commons"-type structure, and that both might have many common solutions, such as
increasing awareness about the problem, introducing legislation, improving technologies,
and much more.

To learn more about archetypes and see examples of other archetypes, we recommend the
work of Peter Senge (e.g., The Fifth Discipline, 1990), Donella Meadows (Thinking in
Systems, 2004: chapter 5) and ​this article written by Daniel Kim. For more readings on
leverage points in systems, we recommend ​this​ article by Donella Meadows.

End of Module Three

This is the end of Module Three. By now we should have a basic understanding of CLDs and
their uses. To recap, we have studied the following topics in this module:

● Causal relationships (positive and negative) between variables

● Different kinds of feedback loops (reinforcing and balancing)
● Behaviour over time graphs
● Archetypes and leverage points

In the next module we will look at more examples and you will further develop your skills in
building your own CLDs.

Further Readings
● Harari, Y.N. (2015). Sapiens : a Brief History of Humankind. New York: Harper.
● Hardin, G. (1968). Tragedy of the Commons. Science 162 (3859), 1243-1248.
● Kim, D.H. (1992). Systems Archetypes I. The Toolbox Reprint Series, Pegasus
Communications, Inc.
● Meadows, D.H. (2008). Thinking in Systems. White River Junction, VT: Chelsea
Green Publishing
● Senge, P. (1990). The Fifth Discipline.
● New York: Currency Doubleday
 Module Four
Building Your Own Systems Models

WASSILY KANDINSKI [CC0 OR PUBLIC DOMAIN], FROM WIKIMEDIA COMMONS

Module Four Overview ​4
Introduction
In the previous module, we learned how to construct and read CLDs, as well as how to
understand some of the phenomena they can help us to recognise, such as feedback loops
and systems archetypes.

However, the process of building your own models from scratch is another challenge, and
that is what we will focus on in this module. To do so, we will model different problems and
describe our thinking throughout the process.

There is always more than one way to model a system, and our process for doing so is not
necessarily the best or only way. Nonetheless, we hope that by walking you through our
thought process for building models, you might learn some good tips for your own model
building. We also hope that the models in this chapter will provoke some thoughts about
some interesting topics!

Module Goals
By the end of this module you should be able to understand:

● How to develop CLD models from scratch

● How to analyse and infer useful insights from CLDs

We will be asking you to develop CLDs for three different topics, and we will provide you with
a certain amount of information for each topic. We will start with a model about a fairly simple
and relatable topic, and then move on to more complex topics.

Module Outline
This module is divided into six sections, each of which is described below:

1. Modelling Procrastination and Productivity

2. Modelling the Utility Death Spiral

3. Modelling the Theory of the Invisible Hand

4. A First Experience of Group Model Building

5. Tips for Model Building

6. Identifying Stocks and Flows

Section 4.1: Modelling Procrastination and Productivity
As mentioned in Module Two, CLDs can be used to represent different perspectives, or
mental models, about a certain issue or topic. In this section we will create two different
CLDs to represent two different mental models about the same topic. The topic we will focus
on is how procrastination affects productivity.

Let's imagine that you interview two different university students about the topic, Mary and
Paul. In response to your question about how procrastination affects productivity, Mary says
the following:

Mary's Mental Model: Procrastination can be useful for productivity

"When I am facing a deadline for an assignment, I usually wait until the last minute to start
working. I used to think that this was a terrible habit, but over the years I've come to realise
that I am simply more productive when under pressure. This is because the closer it gets to
my deadline, and the more remaining tasks there are to complete, the more deadline
pressure I feel. That deadline pressure increases my productivity, such that I complete more
tasks per hour. This means that overall I spend less time on the assignments, with the same
outcome, which I would consider to be a more productive situation!"

Imagine that you then turn to Paul, and he answers your question as follows:

Paul's Mental Model: Procrastination can increase errors and lower productivity

"I share Mary's perspective that when I'm under pressure I complete tasks quicker, but I
would add that when I'm working faster because of time pressure, I make more errors. And
more errors means more remaining tasks to complete. Also, if I don't spot the errors, then
this will reduce the quality of my work. As such I think I am usually better off when I start the
assignment early and complete it with less pressure and thus less errors, rather than waiting
for the deadline pressure to kick in."
Attempting a Model From Scratch
In the next slides we will guide you through a process for developing CLDs of Mary and
Pauls' mental models. However, before we begin, we encourage you to make your own
attempt at developing a CLD of each mental model. You can then compare your models to
ours and see if you have learnt anything from our process. Please note that model
development is a challenging process that gets easier with practice. You are certainly not
expected to be able to produce models quickly and effectively just yet. Nonetheless we
believe that the best way to learn is to try!

Exercise 4A: Identify the Variables

We will begin by creating a CLD of Mary's mental model, shared again below:

"When I am facing a deadline for an assignment, I usually wait until the last minute to start
working. I used to think that this was a terrible habit, but over the years I've come to realise
that I am simply more productive when under pressure. This is because the closer it gets to
my deadline, and the more remaining tasks there are to complete, the more deadline
pressure I feel. That deadline pressure increases my productivity, such that I complete more
tasks per hour. This means that overall I spend less time on the assignments, with the same
outcome, which I would consider to be a more productive situation!"

As a first step in modelling, it can be useful to pick out some of the essential variables that
you will use to build the model. Try to do this yourself before going further. Note that this is
often an iterative process and you may choose quite different variable names when you
begin drawing causal connections. Oftentimes, many modellers will simply come up with the
necessary variables as they build the model. However we think deciding on the variable
names before beginning the model can be a useful exercise for a beginner.

Here are some helpful tips for choosing variables:

● Choose only those variables that are necessary to communicate the essential
message of the model
● Look for relationships and causality, and let them guide you to pick up important
variables. Relationships will involve two or more variables
● Do not include verbs in your variable name; verbs are actions and they are
represented by the causal arrows
● Use clear nouns to represent variable names
● Variables are usually referred to in the text multiple times with different names; try to
group them into one variable name.
Answer to 4A
Below you can see the four variable names that we chose, along with some viable
alternatives for that variable, and some names which we would not consider best practice,
for the reasons described.
You may have also chosen a variable name such as Procrastination. We did not choose to
represent this variable because we think that it is not necessary for our model.

Exercise 4B: Identify the Central Variable(s)

In the previous exercise we have identified four variables that will be used to construct our
CLD. Our next step would be to identify the variable that we consider most important, and
then move from there to draw the necessary causal relations. Please look at the list below
and choose the variable that you believe is most central to our problem.

List of variables:
● Remaining tasks
● Days remaining to finish tasks
● Deadline pressure
● Task completion rate
Answer to 4B
We argue that the Remaining tasks is the central variable to our problem. This is because
Mary's main goal is to reduce the number of Remaining tasks to zero — in other words, to
finish the assignment.

You may have chosen another variable, such as Task completion rate, as the central
variable, arguing that the main issue is productivity. Note that there is no right answer to this
question. Indeed many people might just read the text and start developing a model right
away without thinking of what is the central variable. We are merely suggesting one
approach that we find useful.

Exercise 4C: Identify the Causal Relationships

Once we've identified our central variable, we can move form there to show the cause and
effect relationships that it is a part of.
First we can consider the variable(s) affecting our central variable. What variable do you
think has an effect on the Remaining tasks?
List of remaining variables:

● Days remaining to finish tasks

● Deadline pressure
● Task completion rate
Answer to 4C (1/3)
The more tasks completed per hour/day/week, the less tasks there will be left to do. As such,
the Task completion rate affects the Remaining tasks.

What polarity would you give to this relationship - positive or negative? Go to the next page
to see the answer.
Answer to 4C (2/3)
An increase in the Task completion rate would result in less Remaining tasks, all else equal.
As such, there is a negative relationship from Task completion rate to Remaining tasks.

Next we can identify what variable(s) the Remaining tasks is affecting. What do you think
this variable is affecting in the model? As a reminder, here is some of what Mary said:

"...the more remaining tasks there are to complete, the more deadline pressure I feel."

See how we model this on the next page.

Answer to 4C (3/3)

Based on what Mary has said, more Remaining tasks means more Deadline pressure. As
such there is a positive relationship going from the former to the latter.

Exercise 4D: Finding More Causal Connections

Our next step would be to identify any other variable(s) that are affecting Deadline pressure.

Mary has said that "the closer it gets to my deadline, and the more remaining tasks there are
to complete, the more deadline pressure I feel".

We have already shown how Remaining tasks affects Deadline pressure. However, Mary
has also said that the closer it is to her deadline, the more pressure she feels.

How would you model this? Our answer is on the next page.
Answer to 4D
Fewer Days remaining to finish tasks means more Deadline pressure. As such we draw a
negative relationship from the former to the latter.

Exercise 4E: Completing the Model

What else remains for us to complete the model? Can you finish it yourself?
Answer to 4E
Mary spoke about how Deadline pressure increased her productivity, which we are calling
the Task completion rate in our model. More Deadline pressure means a greater Task
completion rate, and so there is a positive relationship going from the former to the latter.

Can you spot the feedback loop in this model and say whether it is balancing or reinforcing?

There is a balancing loop, which we are calling B1: Pressure means Productivity. This loop
identifies how an increase in Deadline pressure results in an increase in the Task completion
rate, which in turn leads to a reduction in Remaining tasks. This then reduces Deadline
pressure.

Remember, to see all the variables in a loop, simply hover your cursor over the loop name
(in this case, B1).
CLD Development

We so far have used 4 simple steps to develop our CLD diagram of Mary's mental model. To
sum up, we have:

● Identified all variables that can be used to develop our CLD

● Chosen a central variable to start our model development
● Identified causal relationships to and from our central variable
● Completed our diagram by identifying all causal relationships to and from other
variables whilst identifying important feedback loops.

In the next slides we will use the same steps to develop a CLD of Paul's mental model.

Exercise 4F: Paul's Mental Model

Let's try and use the steps on the previous page to create a CLD that represents Paul's
mental model. As a reminder, Paul's answer to the question of how procrastination might
affect productivity is shared again below:

Paul's Mental Model: Procrastination can increase errors and lower productivity

"I share Mary's perspective that when I'm under pressure I complete tasks quicker, but I
would add that deadline pressure causes me to spend less time per task, which increases
the amount of errors I make. More errors means more fixing errors, and so more remaining
tasks to complete. I also sometimes miss these errors before submitting, which lowers the
quality of my work. As such I believe I am better off starting an assignment early and
completing it with less pressure and therefore less errors made."

Try to choose the variable names that you need to build Paul's mental model now. The
variables we chose are on the next slide.

Remember:

● Choose only those variables that are necessary to communicate the essential
message of the model
● Look for relationships and causality, and let them guide you to pick up important
variables. Relationships will involve two or more variables
● Do not include verbs in your variable name; verbs are actions and they are
represented by the causal arrows
● Use clear nouns to represent variable names
● Variables are usually referred to in the text multiple times with different names; try to
group them into one variable name.
Answer to 4F: Identify Causal Relationships

Since Paul says that he agrees with Mary's point about how deadline pressure increases his
productivity, we can keep the same structure as we had for Mary's CLD, and build on top of
it. To the right you will see the new variable names that we have chosen to include in the
model. If you wish to see why we chose these names, what viable alternatives there were,
and what names would not be considered best practice, please click on the icon of each
variable.

Exercise 4G
As a next step, try to draw all the causal connections that Paul mentioned.

Answer to 4G
Paul has said that when he completes tasks faster, he makes more mistakes. As such we
can say that there is a positive relationship going from Task completion rate to Errors. At the
same time, Paul has said that "more errors means more fixing errors, and so more remaining
tasks to complete". We represent this by putting a positive relationship from Errors to
Remaining tasks.

Paul has also said that Errors can lead to a lower Quality of work, given that he might not
spot some errors before submitting the task. This is why we put a negative relationship going
from Errors to Quality of work.
Exercise 4H
Can you spot the new feedback loop that exists in our model? And can you determine
whether it is a reinforcing or a balancing feedback loop?

Answer to 4H
Below we see that there is a reinforcing feedback loop (remember, to see all of the
connections in this loop, simply hover your cursor over the R1 icon).

This is an important loop to note. Whilst B1 shows the mechanism by which procrastination
can increase productivity, R1 shows how it can reduce it. R1 shows how waiting for the
Deadline pressure to kick in can increase your Task completion rate, such that you make
more Errors. This can not only potentially hamper the Quality of work, it can also increase
the Remaining tasks you have, which further increases Deadline pressure and so further
increases the Errors you make! This is a vicious cycle that is good to be aware of. Paul
thinks it is true for him. However maybe Mary thinks differently; maybe she doesn't make
errors under pressure!

Insight: ​Feedback loops help us analyse the underlying structure of a system that creates
the dynamic behaviour we observe.
End of Section 1: Modelling Procrastination and Productivity
The point of the previous section was to guide you through the process of building a CLD.
We thought it was good to start with a model that is somewhat personal and perhaps easy to
relate to. In the next section, we are going to zoom out a bit and look at an issue that is more
macro in nature.

Section 2: Modelling the Utility Death Spiral

It’s generally agreed that the shift to renewable energy is essential if we are to combat global
warming. However, we should still be mindful of how we go about that shift, and any
potential unintended consequences we create along the way. Below is an excerpt from an
article describing the current situation that electric utility companies (companies that produce
and distribute electricity) are facing as a result of the increasing adoption of distributed
generation* technologies such as rooftop solar panels:

"It's a real-world example of the ‘death spiral’ that the industry has so far only considered in
theory: as grid maintenance costs go up and the capital cost of renewable energy moves
down, more customers will be encouraged to leave the grid. In turn, that pushes grid costs
even higher for the remainder of customers, who then have even more incentive to become
self-sufficient."

Before going deeper into the issue by reading the article (linked a few pages ahead), we
challenge you to develop a simple CLD that highlights the cause and effect relationships,
and resulting feedback loop(s), explained in that excerpt.

To makes things clearer, we suggest that you focus your model on how the adoption of
rooftop solar panels (rather than other distributed generation technologies) will affect the
price of electricity for those still on the grid. Try to sketch out your model now and then see
how it compares to ours!

*Distributed generation refers to a variety of technologies that generate electricity at or near

where it will be used, such as solar panels and combined heat and power.

Exercise 4I: Identify the Variables

Try and identify some of the variables you could use to model the death spiral, as described
in the excerpt (shared again below):

"It's a real-world example of the ‘death spiral’ that the industry has so far only considered in
theory: as grid maintenance costs go up and the capital cost of renewable energy moves
down, more customers will be encouraged to leave the grid. In turn, that pushes grid costs
even higher for the remainder of customers, who then have even more incentive to become
self-sufficient."

The variables we initially chose for this model can be seen on the next page.
Answer to 4I
The below is the initial set of variables we chose for our model. Please go to the next page
to see how we initially drew the causal connections.
Exercise 4J: Identify Causal Relationships
Before we go further, can you draw the causal connections between these variables
yourself? Remember to base your model on the below theory:

"...as grid maintenance costs go up and the capital cost of renewable energy moves down,
more customers will be encouraged to leave the grid. In turn, that pushes grid costs even
higher for the remainder of customers, who then have even more incentive to become
self-sufficient."
Answer to 4J (1/3)
"As grid maintenance costs go up... more customers will be encouraged to leave the grid."

As such there is a negative relationship going from Grid maintenance costs to Customers on
the grid. The theory here is that higher costs means higher prices, which means that more
customers will find it economically attractive to leave the grid and generate their own
electricity.

Answer to 4J (2/3)

The excerpt also mentioned that

"as ... the capital cost of renewable energy moves down, more customers will be encouraged
to leave the grid."
This means that as the Cost of solar panels goes down, then there will be fewer Customers
on the grid, because the cost of self-generation has gone down, meaning again that it is
more economically attractive to generate your own electricity (through rooftop solar panels,
for example).

Note that you could have another variable here called Customers leaving the grid, which
would have a negative relationship to Customers on the grid. However we chose to leave
this variable out for simplicity.

Answer to 4J (3/3)

The excerpt has said that as customers leave the grid, "... that pushes grid costs even higher
for the remainder of customers, who then have even more incentive to become
self-sufficient."

As such we drew a negative relationship going from Customers on the grid to Grid
maintenance costs. This is because maintenance costs are relatively fixed costs, that aren't
proportional to the number of people using electricity from the grid. As such, less people on
the grid means that the utility company needs to recover its fixed maintenance costs from
fewer and fewer customers, which they do by increasing prices. This in turn causes more
customers to leave the grid.
This creates a reinforcing loop in the system, which explains why the problem is known as a
death spiral!

Critically Assessing the Model

Developing a CLD is always an iterative process, and we constantly need to reflect on
whether our model is accurately representing what we/someone else thinks. Thus it's
important to critically examine each and every causal link, to ensure that it is accurate.

Upon examining this model, we have noticed an error: a way in which it does not properly
represent the mental model in the excerpt. Can you spot this error?
Exercise 4K
The excerpt has said that as customers leave the grid, "that pushes grid costs even higher
for the remainder of customers.."

Is this accurately represented by the relationship on the right? We believe it is not. How
would you best represent the relationships in a mode? Go to the next page to see how we
chose to model it!

Answer to 4K

Given that the author mentions the costs being greater for the remainder of customers on
the grid, we chose to model the situation as shown here. A decrease in the Customers on
the grid will not decrease the Grid maintenance costs (indeed, it might even reduce it
slightly). Rather, the author is saying that it will affect the Grid costs per customer, as less
customers means less people to split the costs with.
Grid maintenance costs can then be modelled as another factor affecting Grid costs per
customer. The higher the maintenance costs, the higher the costs per customer, so there is
a positive relationship going from the former to the latter.

As such, now our model looks like this.

However, upon further discussion, we decided to make some more changes to the model, to
make it more easily understood and more valuable.
The article seems to assume that people leave the grid because there is a financial incentive
to do so. As such, we decided to add a variable called Incentive to leave the grid because
we felt it made this assumption more explicit and easily understood in the model.

The greater the Grid costs per customer, the greater the customer's Incentive to leave the
grid. At the same time, if the alternative energy source decreases in price (i.e. if Capital cost
of solar panels decreases), then that will also create a greater Incentive to leave the grid.

Insight
CLD development is an iterative process where we always assess and critically challenge
our CLD so as to improve it.
At this stage, we are satisfied that our model accurately represents the mental model
expressed in the excerpt shared earlier.

However, CLDs are not only used to capture the meaning of what someone else has said.
They are also great analytical tools. And as mentioned in Module Two, they can be great for
highlighting assumptions in mental models, and allowing us to critically reflect on those
assumptions.

In this case, we can use the model to discover and highlight the assumptions that are implicit
in the author's mental model about the death spiral.

Before we show how we went about this, are there any assumptions that you notice yourself,
which are implicit in the causal relationships shown in the model? And are there any
changes you would make to the model in order to highlight those assumptions?

We noticed one assumption implicit in the negative relationship going from Incentive to leave
the grid to Customers on the grid. This relationship assumes that households will be aware
of this incentive, i.e., that they will be aware of not only increasing utility bills (which many
people may not check), but also of the decreasing cost of solar panels. If households aren't
aware of this incentive, then it will never cause them to leave the grid!

We chose to highlight this assumption by including an extra variable in our model, known as
Awareness of incentive to leave the grid. As such, our model is now making clear that both
the incentive and the awareness of the incentive will be important in influencing the numbers
of customers on the grid.

We also said that as the incentive increases, so should the awareness, as more and more
solar panel salespeople are calling to your door, and are doing so with a better economic
case for their product!
Another assumption we noticed in the model is that increasing Grid costs per customer leads
to a higher Incentive to leave the grid. It seems that the author of the article is assuming that
an increase in Grid costs per customer will lead to an increased incentive to leave the grid
because utility companies will pass on increased costs to customers, in the form of
increased electricity prices.

We chose to model this assumption by saying that an increase in Grid costs per customer
leads to an increase in the Price of electricity from the grid. An increase in that in turn leads
to an increase in the Incentive to leave the grid.

But do higher grid costs lead to higher prices? Now that we've highlighted the assumption,
it's time to reflect on it.

We did some research to better inform ourselves. We found that, in most US states, the
price of electricity is regulated. Each year, utilities go to a state body, known as the
commissioner, and give them an estimate of the total costs they are likely to bear in
supplying the demand for electricity in a given area. The commissioner then calculates the
expected demand for a given area in the coming year, and uses that to determine the retail
price that utilities will need to charge in order to achieve cost recovery.* So if demand falls
due to more people generating their own electricity, and if costs do not reduce in proportion
to lost revenues, then the commission will likely end up allowing the utility to charge a higher
price for electricity.

*For example, if the expected costs are $10 million, and the expected demand for electricity
is 10 million MWhs, then the commissioner will set the retail price of electricity from the grid
at something like $1.01 per MWh (the extra $.01 per MWh allows for some profit for the utility
and its investors). However if demand drops to 80 MWhs (due to people using alternative
energy sources, for example), and the resulting expected costs of supplying this demand
only drop to $90 million, then the commissioner will likely set the price as $1.125 per MWh
(90 divided by 80, rather than 100 divided by 100).

So now the question becomes: does a reduction in demand for electricity from the grid (due
to rooftop solar use) decrease a utility's costs (through reduced production costs, for
example) such that costs avoided compensate sufficiently for revenues lost?

Some articles we found suggest that if demand for electricity from the grid reduces—due to
rooftop solar use, for example—then the utility's costs do not reduce in proportion to its lost
revenues, because many of the utility's costs are fixed, such as maintenance of the grid.

However other articles we read suggested that utilities are greatly underestimating the
benefits they experience from rooftop solar (and other distributed generation technologies).
Such articles argue that rooftop solar brings many benefits that are not accounted for (for
example, less stress on the grid, as electricity is consumed at the point of production rather
than having to travel via the grid).

As such, they argue that rooftop solar adoption will decrease the utilities' costs in a way that
is roughly proportional to the revenues that they lose as a result of rooftop solar use. This
means that the increasing spread of rooftop solar will not necessitate the increase in the
price of electricity from the grid.
These are all interesting points to note, but our aim here is not to provide an answer. Rather,
we hope to have demonstrated that modelling this problem helped us to ask the right
questions; it helped us to spot the assumptions implicit in the author's mental model and to
critically reflect on those so as to better inform ourselves about the validity of a certain
perspective (in this case, the death spiral hypothesis).

In cases such as these, CLDs can help us to get a better grasp of a problem, and then
communicate our increased understanding (and our assumptions) to others. This CLD could
also be used to analyse certain policies to address the issue of the death spiral, although
that might require expansion of the model, and would likely require us to develop a
quantitative simulation model.

A Note on Determining what is Included in Your Model

As mentioned in Module Two, models are abstractions of reality. As such, they cannot
represent all the details of reality. This means that you must decide on what to include and
what not to include in your model.

There are essentially two things to consider here: the boundary and the resolution of your
model. We will discuss each of these now.

A Model's Boundary

The boundary of a model means the topics/issues that you will be focusing on in your model.
For example, in the previous model on the death spiral, within our boundary was the price of
electricity from the grid, the cost of renewable energy, the number of people on the gird, etc.
Other potentially relevant factors such as the price of fossil fuels were not included. We say
that such variables were outside the boundary of the model. In this case, we considered
these issues to be outside the boundary of our model because although they have an effect
on the speed at which the death spiral may happen, they are not part of the death spiral
hypothesis itself.

Deciding on your model's boundary is always done on pragmatic grounds – it all depends on
what is useful for your model’s purpose. Not including important issues has obvious
downfalls, but including too many simply clutters the diagram and takes attention away from
the most important causal relationships you want to display. In the case of the death spiral
model we decided that including more variables/topics (such as the price of fossil fuels) was
not necessary and would probably have led to a CLD that would be too large to be useful,
especially given that this is an introductory course to CLDs. Indeed, knowing your audience
is another important factor to consider when deciding on a model boundary.
A Model's Resolution
Once your boundary is decided, you need to decide on the Resolution of your model, which
means the level of detail with which you will represent certain topics/issues. A model with a
higher resolution will show more of the causal mechanisms that link variables together,
whereas a model with lower resolution will not give as much detail about the causal
mechanisms in the system at hand.
For example, in the death spiral model, we would say that we increased the resolution for
the model when we included the Price of electricity from the grid variable in between the Grid
costs per customer and the Incentive to leave the grid. Showing the assumed causal
mechanisms in greater detail here (i.e. increasing the resolution) helped us to better highlight
and discuss the assumptions in the death spiral hypothesis. In this case, it highlighted how
we assumed that higher Grid costs per customer would lead to a higher Price of electricity
from the grid, which would then increase the Incentive to leave the grid.
If, on the other hand, you felt an increase in the Grid costs per customer would obviously
lead to a greater Incentive to leave the grid, then perhaps you wouldn't need to include the
Price of electricity variable. However we felt that this relationship was not so obvious and so
was worth highlighting,
Lower resolution (assumptions less explicit):

Higher resolution (assumptions are more explicitly represented):

Again, deciding on the resolution of your model is done on pragmatic grounds: we want to
avoid models that have too many variables, but we also want to include sufficient detail for
representing important assumptions and causal mechanisms.

Determining a Model's Boundary and Resolution

Ultimately, there are no hard rules for determining the boundary and resolution that you set
for your model. It is a decision to be made on pragmatic grounds, and it will depend on many
factors, such as the reason for which you are making the CLD, and the audience you are
creating it for.

Also, it's important to remember that determining the model's boundary and resolution is an
iterative process, and we encourage you to consistently reflect on whether your CLD's
boundary is too narrow/broad, and its resolution too high/low.

Keep this in mind as we move to the next section.

End of Section 2: Modelling the Utility Death Spiral

That brings us to the end of section two. We hope that by now you have learned a bit more
about how to create a CLD that accurately represents someone's mental model/hypothesis,
and that you are aware of some of the important considerations to be made when doing so.
We discussed the importance of deciding on a boundary and level of resolution for your
model, as well as the importance of constantly reflecting on your model and iteratively
improving it. We hope to have also demonstrated how building a CLD of someone's
hypothesis can help one to critically reflect on that hypothesis in a deeper way than we
usually do.
Section 3: Modelling the Theory of the Invisible Hand

In this section, we are going to make things a little more challenging. Rather than provide
you with text, we are asking you to watch a video and build a CLD of the theory it describes.
The theory is Adam Smith's beliefs about how the price of commodities can be kept in
balance when economic actors are allowed to act in their own self-interest. This theory has
become known as the Theory of the Invisible Hand.

Please click on the video below to see an explanation of Smith's theory.

Exercise 4L: Modelling the Theory

We challenge you to build a CLD that succinctly explains the cause and effect relationships
assumed in Adam Smith's theory.

We will begin presenting the model we built on the next page. Our model describes the
theory using the example in the video of how different bakers of scones regulate the price of
scones whilst acting in their own self-interest. We decided to keep it simple and focus just on
this aspect, leaving the supply chains for scones outside the boundary of our model.
Answer to 4L
We began our model with a variable representing the Price of scones from baker A. We then
asked ourselves what affects this variable.

Smith assumes that when there is an increase in demand for something, sellers will increase
the price of their product, all else equal. As such, we model a positive relationship going from
Demand for scones from baker A to Price of scones from baker A.
However demand alone will not determine the price of a commodity. According to the theory
known as the Law of Supply and Demand, both supply and demand will impact the price of a
good. The theory says that when demand for a product is high and supply is relatively low,
then prices will be higher. The theory also suggests that when demand is low and supply is
relatively high, the prices will be lower.

For example, if the baker knows he will have 20 people that want to buy his scones, but he
only has 15 left, he or she might increase the price of the scones. Alternatively, if the baker
makes too many scones in a given day, and has many left in the shop, then he or she might
consider lowering the prices of scones (offering a two-for-one promotion, for example).

Next we can ask ourselves what determines the supply of scones in the first place. Smith's
theory was about people acting in their own self-interest. As such, we can assume that the
baker will only supply scones if there is a profit to be made in doing so. The greater the
Potential profit for baker A, the greater the Supply of scones from baker A, as shown in our
model.
The potential profit will increase when the price is higher, all else equal, and so we included
a positive relationship from Price of scones from baker A to Potential profit for baker A.

We also decided to include a variable called the Cost of making one scone, which can be
thought of as the cost of ingredients, labor, and all other factors needed to make one scone.
The higher the cost, the lower the potential profit, all else equal.
As such, we see the first balancing loop that explains part of the dynamics of Smith's theory.
This loop says that the more profit there is to be made from supplying a good, the more of it
will be supplied. And the more of it supplied, the lower the price, and so the lower the
potential profit, and the lower the supply.
This loop helps to regulate the price of scones, and brings supply in line with demand.
However, the theory doesn't end there. Smith also talks about how competition between
different actors also helps regulate the price of goods. In the video, we saw how another
baker noticed the potential profit to be made in supplying scones. As such, the dynamics for
baker B can be modelled in the same way, as shown.
Now that there are two bakers supplying scones, there is competition in the market. This
means that customers have a choice as to which baker to go to for scones.

To model this, we said that the prices of scones from each baker will determine the Relative
value of scones from baker A: a higher price from baker A leads to a lower relative value,
whereas a higher price from baker B leads to a higher Relative value of scones from baker
A. Given that price is the only factor affecting the relative value of scones from each baker,
we are of course implicitly assuming (for simplification) that the quality of scones from each
baker is the same.
The Relative value of scones from baker A will then determine the demand for scones from
each baker, as customers choose what they believe to be the best value. Note that this
relationship in itself assumes that customers are aware of the price of scones from both
bakers, which is not the case in all markets!
Now we see an additional balancing loop for each baker. Loop B2 says that as the Relative
value of scones from baker A increases, so will the Demand for scones from baker A. This
will in turn increase the Price of scones from baker A, which reduces the Relative value of
scones from baker A.

At the same time, an increase in the Relative value of scones from baker A will cause baker
B's demand to drop, which causes their price to drop, which in turn decreases the Relative
value of scones from baker A, or in other words increases their relative value. As such, loops
B2 and B3 describe the way in which competition between bakers regulates the prices that
they each charge. When one baker increases their price, the other becomes more popular,
causing the first baker to bring their price back down so as to remain competitive.

At the same time, neither baker is likely to lower the price of their scones so much that they
no longer make a profit, as described by loops B4 and B1.

End of Section 3: Modelling the Theory of the Invisible Hand

We hope that in this section we have shown that CLDs can help us to better understand
abstract theories by explicitly representing each of the cause and effect relationships that
they assume. We appreciate that this was not an easy modelling exercise, but we felt that it
was a good idea to give you a challenge!
Section 4: A First Experience of Group Model Building
In this section we will be giving you a challenge to complete with a group of people. This can
be any group: your classmates, your colleagues, your friends, or even your family! The
purpose of this exercise is to familiarise yourself with what is known as group model building
(GMB), a methodology within system dynamics that we briefly introduced in Module Two.

As mentioned there, GMB involves engaging different stakeholders to a problem in a

process of jointly building a model. The aim is to integrate different stakeholder perspectives,
and allow for the development of a shared understanding of a problem. The process acts as
an opportunity to improve communication between stakeholders, and the synthesis of
different perspectives on a problem can often lead to the discovery of new solutions.
Additionally, by involving stakeholders in the analysis, the group is more likely to be
supportive of the outcomes of that analysis (i.e., the decision, policy, or strategy developed),
which is extremely valuable in the implementation phase.

A Typical GMB Session

A typical GMB session involves identifying and gathering between 5-15 stakeholders in a
room, together with at least two people experienced in system dynamics and facilitation. One
of these system dynamicists will facilitate the group process, while the other will focus on
building the model based on the group's conversation. The modeller can also interact with
the group when necessary, for example to check that the model represents what the group
has been saying. This allows the facilitator to keep her/his attention focused on the group
dynamics, and allows the modeller to focus on producing an accurate and clear model that
everyone can read.

The group usually starts by building a CLD of a problem, and may at a later stage create a
simulation model if that seems necessary. It may take several half-day sessions until a group
has constructed a model they can agree on, and from there it might take several more
sessions to use the model for analysing potential solutions to the problem. In the end, the
hope is that the group develops a shared understanding and a common motivation for
implementing a certain solution/strategy.

Simulating a GMB Session

To get an idea of what running a GMB session is like, it is best to be involved in a real
situation in which you and those around you are strongly invested in the outcome of the
analysis. However, there is no harm in using a little imagination and running an exercise in
which we imagine ourselves as stakeholders to a problem - which is what we will do in this
section! We will use a real-life case of GMB that The Sustainability Lab were involved in, and
will ask you and your group to imagine yourselves as different actors in this complex
situation.

Before we explain more about what you will do in this exercise, it will be helpful to get a bit of
background information.
The Martina Bustos Community
The case we will deal with occurred in 2014 and 2015, when fellows of The Lab's Global
Sustainability Fellowship program were working together with the Martina Bustos community.
This is a settlement that is located 4 km from downtown Liberia, the capital of the
Guanacaste region in Costa Rica. The settlement was established about 15-20 years ago
when a private landowner, named Martina Bustos, donated the territory so that low-income
families could have a place to live. However, Martina never formally handed over the land
titles, and to this day they remain with her children.

Nonetheless, many poor migrant families have moved to the area, and the population on the
settlement has grown to between 3000-3500 people, mostly of Nicaraguan descent. The
lack of land titles has caused many problems as, without them, the community is not formally
recognised by the Costa Rican government. As a result, there is no legal obligation for the
government to provide access to basic services such as electricity, waste management, and
clean water.

In the summers of 2014 and 2015, participants in the Sustainability Lab’s Global
Sustainability Fellowship program began working with the Martina Bustos community. The
purpose of the work was to create a community development plan that would serve as the
blueprint for creating a more robust and sustainable community. The fellows worked with
members of the community (mainly the women’s association) to identify central problems as
perceived by the community members, and try to find a strategy for resolving them. Much
work had already been done by other NGOs in this area, but The Sustainability Lab’s
approach was unique in that it took a holistic/systems perspective on the community,
addressing multiple problems at once (economic, social, environmental, and motivational)
rather than focusing on just one issue at a time (for example, just housing or just water).

The Use of CLDs

As part of this work, the fellows held GMB sessions with the community members, the
purpose of which was to discover together how the problems of the community might be
interlinked. This analysis would then aid in the development of a holistic strategy for
addressing the community's issues. For these sessions, about ten community members
were invited to the local university. They participated in a two-hour session which was
facilitated in Spanish by Vanessa Armendariz, one of the fellows.
Role-Playing

For this exercise, we want you and your group to split up into different roles, each of which is
described below. If the group is larger than 12 people, then we suggest that you form two
groups and conduct the exercise separately.

Role 1: Members of Martina Bustos (2 to 10 people): Your task is to put yourself in the shoes
of members of the community. To become familiar with the nature of the issues faced by the
community, we will provide you with some quotes from the community members (shared a
few pages ahead). To keep things simple, you should try focus only on the issues in those
quotes, so that the scope of the model does not become too broad to handle. Your job is
essentially to experience a GMB session as if you were a participant, and we hope that this
will be enlightening for when you become a facilitator or modeller.

Role 2: The Facilitator: You are essentially leading the group model building session. Your
role is to facilitate the process from start to end, managing group dynamics and ensuring that
all participants are able to express their views, and that the conversation maintains its focus
on the issues at hand.

Role 3: The Modeller: Your role is to create a CLD based on what the members of the
community are saying. While the facilitator will do most of the interacting with the group, you
may interject every now and then when you need to clarify what someone has been saying,
and whether your model is reflecting that accurately.

Logistics
It will help if you can do this exercise in a room with a whiteboard upon which you can draw
the model, but if that is not possible then you can use a large piece of paper that the
modeller can easily show to the other participants. We suggest that the facilitator stands in
front of the group, and arranges the others in a semi-circle in front of her/him, as this creates
the most balanced way for a group to interact. Ideally, the modeller would stand behind the
facilitator, at the white board. Alternatively, if the modeller chooses to build the model on a
computer then we suggest that they use a projector/digital display to make their model easily
visible to the whole group.
Focus of the Session
To keep things simple in this exercise, we can imagine that the agenda for the meeting has
been set around looking at issues related to the community's lack of land titles. We suggest
that those playing the role of community members maintain their focus on just that issue, to
avoid too many topics being discussed at once.

One of the facilitator's responsibilities is to keep the discussion focused, and if one particular
issue has been agreed upon before the session, then that should help in determining what is
off and on topic. Keeping the focus on an agreed boundary will allow for a more productive
session, rather than letting the group's discussion touch upon many topics without delving
into any in depth.

Of course, you and your group are unlikely to know much about the land titles issue of the
Martina Bustos community. To help you play their role, you will find a number of quotes from
participants in the actual GMB session on the next page. You can take inspiration from these
quotes to try to put yourself in the shoes of the community members, and see the problem
from their perspective. You can then relate these issues to the facilitator, who together with
the modeller will try and develop a CLD that combines together the different perspectives.

We suggest that the people playing the role of facilitator and modeller do not look at these
quotes, as part of the challenge of facilitating a GMB session is dealing with and
understanding the issues expressed by participants as they arise, without any prior
knowledge of them. So if you are playing the role of the facilitator or modeller, please go
ahead and skip the next page and move to exercise 4M.

Quotes from the Community Members

Below is a series of quotes from the Martina Bustos community members during their GMB
session with us in 2015. Please note that these have been translated from Spanish and
simplified for educational purposes. For confidentiality purposes, we have not specified the
author of each quote.

"I don't really feel like I belong in this community. To feel belonging, you need to feel a sense
of stability in a place. Without the land titles, I feel that I could be kicked out of my home at
any time. I think that affects how we see each other here: the lack of land titles means there
is no feeling of community."

"At the moment, there is too much conflict in the community for progress to happen. Only
when we work together will we achieve any real improvements."

"I think we need the landowner to hand over the land titles before the people here commit to
improving the infrastructure in the community. Without the titles, there is no sense of
community, and without the sense of community, there are no successful projects carried
through."
"The landowner has said time and time again that he does not think we are ready to have
the land titles. He says that until we act like a community, he will not hand over the titles."

"The government have said they will not support us with funding until we have demonstrated
that we can work together on projects and see them through to the end."

Exercise 4M: Conducting the Session

Now that you have the background and can understand some of the perspectives held by
the community members in relation to the land titles issue, it's time to try to conduct the GMB
session. Once everyone is gathered together in a room, the facilitator can begin by
explaining what the focus of the session will be (in this case, the land titles issue).

In some GMB sessions, the participants may not be familiar with CLDs, and so the notation
and meaning of causal relationships would have to be explained first.

Once everyone is clear on the focus of the session, and on how CLDs work, you can begin
the modelling. There are many ways to being a GMB session. A list of 'scripts' for how to
carry out GMB sessions can be found in this free online resource.

For this session, we recommend that you begin the modelling by drawing up the central
variable on the board, which would be 'Access to land titles'. As in the previous exercises in
this module, you can work from here to identify what affects and is affected by this variable.
The only different is that you are now doing this exercise as a group, rather than an
individual. This can be considerably more challenging, but it may also lead to more fruitful
insights, as different stakeholders share different knowledge or perspectives about the
problem.

As the group begin identifying causal relationships, the modeller should begin drawing these
on the board/computer. When necessary, the modeller should clarify whether she/he is
modelling the thoughts of the participants correctly. The modeller should also look out for
feedback loops in the model. When one is identified, the facilitator can ask the group to
discuss whether this is a reinforcing or balancing loop. If it is reinforcing, the facilitator should
ask whether it is currently working in a desirable or undesirable direction. This exercise can
help participants move from linear to non-linear reasoning and insights about the system at
hand.
Exercise 4N: Identifying Solutions

Once you have constructed a model of the land titles issue, the facilitator could begin to ask
the group to use the model to identify leverage points, i.e. variables on which they have
control and which can lead to a desirable change in the system (in this case, the actions that
could help in getting the land titles). It's important that the facilitator remains neutral in this
and at all stages of the session. The facilitator's role is to manage group dynamics and to
keep the discussion within the range of focus, but they should not let their own opinions
influence the group's decisions or beliefs.

Typically, once a group has identified potential solutions to the issue at hand, the facilitator
can conclude by reflecting on what has been done in the session, and what the decided
outcomes were (in the case of Martina Bustos, the group focused on identifying how to
resolve the land titles issue). Sometimes the outcome of a GMB session can be purely
educational, but ideally a GMB session will lead the group to identify concrete steps or
strategies that they can enact in order to resolve a common problem.

There is no right or wrong way to model the issues that the Martina Bustos community spoke
of. Nonetheless, if you are interested in seeing the way we chose to model it, then you can
see the model in section three of the next module, Module Five, which focuses on case
studies in which CLDs were used.

Trying Another GMB Session

Having completed this session, you may wish to build on your experience and conduct
another GMB session with your group, this time on an issue of your choosing. Ideally, the
topic you choose would be something that everyone is interested and invested in, and
something that would be interesting to study using system dynamics. This would ideally be
something that involves a complex system with many interacting parts and the potential for
unintended consequences.

End of Section 4: A First Experience of Group Model Building

Facilitating a GMB session is not an easy task, nor is gathering a group of people to
participate in one. Nonetheless, both the facilitation and the gathering of people are real life
challenges that you may face should you decide to practice system dynamics, and so we
wanted to give you an opportunity to challenge yourself in this way. If things did not go
smoothly, or as planned, that is more than understandable and part of the learning process.
The goal is to achieve incremental improvements every time you facilitate, or even
participate, in a GMB session.
Section 5: Tips for Model Building
The previous sections have led us through several cases where you have built your own
models. We hope that by now you are beginning to feel comfortable in constructing your own
causal loop diagrams. In this section we will step back and reflect on some of the best
practices for model building.

Modelling Tips

As we mentioned at the beginning of this module, there are many different processes by
which a model can be built, and our recommended process is just one. As you gain
experience building models, you are likely to begin doing so naturally, with less focus on the
process and more focus on the issue.
Nonetheless, you may find the below list a useful resource as a rough step by step guide for
one way of developing a CLD:
● Collect your data (i.e. do interviews with relevant stakeholders/experts, read articles,
collective quantitative data, etc.)*.
● Determine the boundary of your model: what issues you will include, and what ones
you will leave out. Note that this is an iterative process.
● Determine the resolution of your model: decide on the right level of abstraction/detail.
This is also an iterative process.
● Start your diagram from your central variable (i.e. the most problematic or important
variable).
● Find out what affects and is affected by your central variable, keeping in mind the
boundaries that you have set.
● Draw the correct causal links to and from the central variable, denoting whether it is a
positive or a negative relationship.
● Highlight significant delays between causes and effects by using a dashed arrow for
such relationships.
● Look for and identify important feedback loops in your model, and denote them as
either balancing or reinforcing. You may also want to give certain feedback loops a
name.
After development we recommend that you:
● Challenge and critically assess the causal links and model boundary/resolution and
generally seek to improve your model by iteration.
● Try and determine whether the reinforcing loops in your model are currently acting as
a virtuous cycle or as a vicious cycle
● Determine whether the balancing loops are hindering progress or are giving a
desired stability to the system
● Identify the major assumptions in your model and make these assumptions clear to
others (for example if you write a paper about a CLD, we recommend having a
section called "model assumptions" which explains each assumption and ideally
gives evidence to support each one)
 ● Create a dynamic story of the current problem, unfolding your model step by step

Regarding the layout of variables on a page/screen, we recommend that you:

● Try to make loops look like loops! In other words, use curved lines for feedback loops
so that they look like circles or ovals on the page, making them more easily
identifiable
● Organize your diagrams to minimize crossed lines.

Perhaps the most important part of the model building process is to express any insights or
clarity you have gained as a result of the process. This kind of reflection can help you to see
the value and the limitations of CLDs.

*The information on which you base your CLD can come from many sources. This includes
surveys, organisational databases, news papers, research papers, government publications,
TV interviews, industry statistics, video recordings, transcripts, personal interviews, personal
observations, structured observation, analysis and experimentation, and much more.

When Can We Use CLDs?

Before we conclude this module, it's important that we reflect a bit on the use of CLDs. Of
course, CLDs are useful tools in many contexts, but not all. As the saying goes, having a
hammer does not mean that everything is a nail!

So when are CLDs likely to be useful?

There is no hard and fast rule for this, as it depends on judgement. However, below are
some scenarios in which CLDs have proven useful:

In analysing problems where a system or variable within the system is changing or

suspected to eventually change in an undesirable way. For example, if the number of
high-school drop outs is increasing, the inequality in a country is on the rise, or if a
population of a specific species is declining, CLDs can be useful in developing and
communicating an understanding of why.
CLDs can also be used when a group of people hold differing views on what is causing a
certain problem, and hold different knowledge about that problem which requires synthesis
in order to bring about a more holistic understanding. For example, people from different
departments in a business might hold different views about why the business is losing
market share, and a CLD can be a very useful tool for synthesising these different
perspectives and improving the likelihood of finding a successful solution. It also serves as a
great communication tool between different stakeholders in such settings.
Limitations of CLDs

But what then, are the limitations of CLDs? We briefly discussed this in Module Two, but it
may be worth revisiting now that you have a deeper understanding of what CLDs are.

Firstly, as systems thinkers we are interested in understanding how things change over time.
CLDs, however, hold a limited ability to provide an understanding of the behavioural output
of a system. Without quantifying your model, it is difficult to infer the magnitude of change in
different variables over time, what feedback loops are dominating the system at any given
moment, and when shifts in feedback dominance are likely to occur. In order to better
understand the behaviour created by the system structure, a formal simulation model is
needed.

Another limitation of CLDs is that it is sometimes difficult to balance a need to simplify with a
need to be precise. When simplifying in order to shed light on a specific issue or part of the
system, the underlying rationale or the assumptions behind the causal links included in the
CLD can become somewhat difficult to grasp. This can be particularly true when one has not
been involved in the creation of the CLD. Whilst the meaning of all variables might be clear
to those who created the model, the meaning to those reading about it may not be so clear.
This limitation can also be true for simulation models.

For further readings on this topic, ​see this paper by David Lane.

End of Section 5: Tips for Model Building

That brings us to the end of section 5. This offered us a chance to reflect on some of the
practices we have learned so far, and on the reasons for which we use CLDs in the first
place, as well as some of their limitations.

In the next section we will introduce a new concept from system dynamics: that of stocks
and flows.
Section 6: Identifying Stocks and Flows

Up until this point in the course, we have built CLDs where there is no distinction between
different types of variables. However, system dynamicists would argue that it is often
important to make a clear distinction between two types of variables: stock variables (simply
known as stocks), and flow variables (known as flows). In this section we will discuss the
difference between stocks and flows, and why it is so important to differentiate between
them.

We will also demonstrate the notation used to highlight and differentiate stocks and flows in
a system dynamics model. Differentiating between stocks and flows is typically done in
simulation models rather than in CLDs, but some CLDs also make the distinction. The
importance of making this distinction should be clear by the end of this section.

Exercise 4O: The Department Store Task

To begin this section, it will be worthwhile to do a small exercise, developed by professor

John Sterman of MIT. The exercise asks you to review the below graph, which shows the
number of people entering and leaving a department store over a 30-minute period, and then
answer the questions found below. If you believe the answer can't be determined, then
simply leave the answer blank.

Questions
1. During which minute did most people enter the store?
2. During which minute did most people leave the store?
3. During which minute were the most people in the store?
 4. During which minute were the fewest people in the store?

Answers to 4O

The answers are as follows:

1. Most people entered the store at minute 4.
2. Most people left the store at minute 21.
3. The most people were in the store during minute 13.
4. The least people were in the store during minute 30.

Reflecting on the Exercise

You may have answered Questions 1 and 2 correctly. ​In this study conducted on MIT
students, 96% of participants answered Question 1 (about when most people entered the
store) correctly, and 95% of participants also answered Question 2 (about when most people
left the store) correctly.

However, you may have had more trouble with questions 3 and 4 (about when the most and
least people are in the store, respectively). In the above-mentioned study, only 44% of
respondents answered Question 3 correctly, and just 31% of respondents answered
Question 4 correctly.

For now, let's focus on the responses to Question 3. 29% of respondents to this question
incorrectly asserted that most people are in the store at minute 8, which is the minute during
which there is the greatest difference between the people entering and leaving the store.
Many people incorrectly assume that this means that this will be the minute during which the
most people will be in the store. In reality, the number of people in the store will continue to
grow for as long as there are more people entering the store than leaving the store. For that
reason, we know that most people are in the store at minute 21, as up until then the number
of people entering the store is greater than the number of people leaving. After minute 21,
there are more people leaving than entering, which means that the number of people in the
store starts to decrease.

Accumulation
The reason many people fail on questions 3 is because they fail to understand the principles
of accumulation. In system dynamics terms, we would say that they fail to distinguish
between stocks and flows. Flows are variables that happen over time: for example, the
amount of water flowing from a tap per second is a flow, and the water flowing down the
drain per second is also a flow. Stocks refer to variables that accumulate flows. For example,
the level of water in a bathtub is a stock that accumulates the flow of water into and out of
the bathtub. The water flowing into the bathtub per second would be known as an inflow to
that stock (because it adds to the stock) whereas the water flowing out of the bathtub per
second would be known as an outflow from that stock (because it subtracts from it).

Similarly, the number of people in a department store is a stock, and it accumulates the
flows of people entering the store per minute (the inflow), and people leaving the store per
minute (the outflow). The important thing to remember about stocks and flows is that a stock
will increase if and only if its inflow(s) is greater than its outflow(s). So, the number of people
in a department store will always increase when the people entering the store is greater than
the people leaving the store.

Notation for Stocks and Flows

Typically, CLDs do not distinguish between stock and flow variables. For example, in the
department store task, a simple CLD of the situation might look like this:

However, if we wanted to highlight the stocks and flows in this model, then we would use
notation as seen below:
The rectangular box around the stock represents the fact that this is a variable where things
accumulate or drain away from. The hourglass symbol for the flows symbolise the fact that
they always happen over time (per second, per minute, per month, per year). Note that the
inflow (people entering the store per minute) is pointing into the stock, whereas the outflow
(people leaving the store per minute) is coming out of the stock. This represents the fact that
inflows add to a stock, whereas outflows subtract from a stock.

Determine Stocks from Flows

So the easiest way to determine if a variable is a stock or a flow is to ask yourself whether
the variable happens over time (i.e. per second, per week, per year). If it does, then it is a
flow. And if the variable is something that accumulates flows, then it is a stock.

Exercise 4P: Recognising Stocks and Flows

The distinction between stocks and flows might not be immediately obvious to you. However
it should become clearer when we begin to consider more and more examples of what a
stock is, and what a flow is.
Below you can see a list of variables. Can you spot the stocks and the flows? Some of these
stocks and flows are causally related. Can you also draw the stock and flow relationships
between these variables, using the notation presented on the previous page? Remember
that inflows are arrows pointing into the stock, and outflows are arrows pointing out from the
stock.
The answer is on the next page.
Variables
● Water in a bathtub
● Litres of water flowing into a bathtub per second
● Litres of water flowing out of a bathtub per second
● World population
● Deaths per year
● Births per year
● Money in a bank account
● Dollars withdrawn from a bank account per month
● Dollars deposited into a bank account per month
● Goods in a warehouse
● Products delivered to a warehouse per day
● Products shipped from a warehouse per day
 ● Houses in a neighbourhood
● Houses constructed per month
● Houses demolished per month
● Trees in a forest
● Trees dying/cut per month
● Trees grown/planted per month

Answers to 4P
Did you recognise the stocks from the flows? And did you draw the draw the inflows and
outflows correctly?
Not All Stocks Are Material

Note that stocks do not always have to refer to physical things. Intangible variables can also
be stocks. For example, if you were building a model about the reputation of a certain brand,
you might highlight the fact that reputation is something that accumulates over time. As such
you might model it as a stock, which is affected by flows such as "positive experiences with
the brand per month" (an inflow), and "negative experiences with the brand" (an outflow).

The Importance of Distinguishing Stocks from Flows

The research on the department store task demonstrates that many people have trouble with
the principles of accumulation, i.e. with distinguishing stocks from flows. In the context of
determining the number of people in a store at any given time, the common inability to
recognise stocks form flows is not a life or death situation. Nonetheless, this cognitive blind
spot can have important consequences. For example, many studies have shown that people
don’t understand this fundamental characteristic of the greenhouse gas (GHG) problem (for
example, see this study). Indeed, it is often assumed that if we stabilise our GHG emissions,
then the stock of GHGs in the atmosphere will also stabilise. But is this true? Given that our
current rate of GHG emissions is far above the rate at which GHGs are absorbed by natural
sinks, what do you think will happen if we stabilise emissions at their current rate?

Write down your answer now, and we'll come back to it after doing the next exercise.

Exercise 4Q

To help you think through this, try to draw a stock and flow model of how the level of CO2
(the main greenhouse gas) in the atmosphere changes. To keep things simple, you should
consider just three variables:
● The CO2 emitted into the atmosphere per year
● The amount of CO2 in the atmosphere
● The CO2 absorbed from the atmosphere each year (through carbon capture from
trees and the seas, for example).
The answer is on the next page.

Answer to 4Q (1/3)
The amount of CO2 in the atmosphere is a stock, like the water in the bathtub.
Answer to 4Q (2/3)

The annual global CO2 emissions is the inflow to this stock, like the water flowing through
the tap into the bathtub.

Answer to 4Q (3/3)
Finally, the annual CO2 absorbed from the atmosphere is the outflow, like the water flowing
down the drain.

This removal of CO2 from the atmosphere happens through natural processes. For example,
when trees photosynthesise, they take CO2 out of the atmosphere and use it to grow.
Ecosystems that take carbon out of the atmosphere are known as carbon sinks.

Did you correctly distinguish the stock from the flows, and the inflow from the outflow?

Exercise 4R
Given that we now understand the GHG problem in terms of stocks and flows, what do you
think would happen if we stabilised CO2 emissions at their current rate, given that they are
currently far higher than the rate of CO2 being absorbed from the atmosphere? Would the
stock of CO2 in the atmosphere continue to grow, stabilise, or go into decline? Has building
the stock and flow model of the situation has helped you think through this issue in any way?

Answer to 4R
If we stabilise emissions at current rates, then the CO2 in the atmosphere will not stabilise,
but will continue to grow. In order to stabilise the level of emissions in the atmosphere, it is
necessary to reduce the rate of emissions to the rate of CO2 being absorbed. In other
words, we would need to reduce the inflow to the same level as the outflow. In order to
reduce the level of CO2 in the atmosphere, we would need to bring the emissions to lower
than the rate of absorption.

This can be summarised in the below graph, which displays CO2 concentration levels in the
atmosphere under different emission scenarios.
Visualising the GHG Problem
To understand how flows affect stocks, it can often help to visualise it. The simulator
available on this webpage* is based on a quantified system dynamics model and allows you
to see how different CO2 emission pathways will affect the stock of CO2 in the atmosphere.
Besides increasing our understanding of the greenhouse gas problem, this tool also offers a
great way to visualise how stock and flow systems work. Try and run the simulation under
different emission scenarios and see how that affects the level of CO2 in the atmosphere
over time.

*If the simulator doesn't work at first, try using another browser and/or check your Adobe
Flash Player settings.

End of Section 6: Identifying Stocks and Flows

That brings us to the end of section 6. Being able to differentiate between stocks and flows is
essential when building simulation models, but it is not necessary when creating CLDs.
Nonetheless, as we hoped to have shown, the distinction is still an important one in general
and can help improve reasoning about complex issues – such as climate change and the
accumulation of GHG emissions in the atmosphere
End of Module Four
That brings this module to a close. To recap, we began with an exercise building a model of
a fairly simple and relatable topic, based on an imagined interview with two university
students. From here, we moved to modelling a more complex topic (the utility death spiral),
where we hope to have shown how causal loop diagramming can help in both thinking more
clearly and analytically about an issue, and in identifying the assumptions implicit in
someone's mental model. The third exercise, on the Theory of the Invisible Hand, provided a
greater challenge in asking you to build a CLD based on a video, and also demonstrated
how CLDs can be used to analyse and hopefully better understand abstract theories. In
section four we asked you to imagine yourselves in a real-life group model building situation,
a challenging task no doubt. In section five, we reflected on the best practices for creating
CLDs, and touched again on its limitations. In section 6, we introduced the concepts of
stocks and flows and discussed why it is so important to differentiate between them.

Further readings
Sterman, J. D., & Sterman, J. D. (2014). Business dynamics: Systems thinking and modeling
for a complex world. Boston: Irwin. (we recommend chapters 1, 3, and 5)

Homer, J. B. (1996), Why we iterate: scientific modeling in theory and practice. Syst. Dyn.
Rev., 12: 1-19. doi:10.1002/(SICI)1099-1727(199621)12:1<1::AID-SDR93>3.0.CO;2-P

Homer, J. and Oliva, R. (2001), Maps and models in system dynamics: a response to Coyle.
Syst. Dyn. Rev., 17: 347-355. doi:10.1002/sdr.224

Kim, D. H., & Kim, D. H. (1995). Guidelines for drawing causal loop diagrams. Waltham, MA:
Pegasus Communications.

Sterman, J. (n.d.). Fone Tuning Your Casual Loop Diagram - Part 1. Retrieved from:
https://thesystemsthinker.com/fine-tuning-your-causal-loop-diagrams-part-i/

Rushing, W. (n.d.). Causal Loop Diagrams: Little Known Analytical Tool. Retrieved from
https://www.isixsigma.com/tools-templates/cause-effect/causal-loop-diagrams-little-known-a
nalytical-tool/
 Module Five
Case Studies

WASSILY KANDINSKI [CC0 OR PUBLIC DOMAIN], FROM WIKIMEDIA COMMONS

Module Five Overview ​5
Introduction
Module Five discusses five case studies that demonstrate how CLDs have been used in
certain situations, and the value that they provided. This is then followed by some final
thoughts and a final exercise to conclude the course.

The previous module walked you through the process of building models relating to different
problems. In this module the focus will be on real life case studies in which causal loop
diagrams were used to help analyse a certain problem/highlight a certain issue.

The real value of the process of causal loop diagramming cannot be fully appreciated by
analyzing case studies alone. Oftentimes, one needs to be experiencing the complexity of
the situation firsthand in order to see how modelling can help you or the group you are a part
of to communicate and think clearly, to ask the right questions, and to feel a sense of
comfort that you are able to manage the complexity of the situation without shying away from
it. Nonetheless, the following case studies should provide a sense of how CLDs can help in
certain situations, and will also give you some idea of the wide variety of contexts in which
they have been used.

Module Outline
We will deal with five case studies which are quite different in nature. We hope that by the
end of this chapter you will have a better understanding of how causal loop diagrams have
been used historically, and how you yourself might use them to help tackle a challenge in
your own life/organisation/society.

The five case studies are as described below:

1. Gender Equality in Dutch Technical Universities (Inge Bleijenbergh and Marloes Van
Engen)

2. Hewlett-Packard Market Share (Deborah Campbell)

3. The Martina Bustos Community (The Sustainability Laboratory)

4. The ISLAND project (Andrea Bassi, Gilbert Probst, & Prakash Deenapanray)

5. Building a Healthcare System (Markus Schwaninger)

Case study 1: Gender Equality in Dutch Technical Universities

The Context
Gender inequality is a complex problem. Indeed, stakeholders often disagree on the
existence of the problem itself, its causes, or its solutions. The problem is particularly evident
in Dutch technical universities, which have a very low ratio of female to male professors. In
addition, the progress towards an equal ratio is considered slow.

The case study we discuss here was performed in the Social Sciences and Humanities
University and the Technical University in the Netherlands. This case occurred in 2009,
when professors Inge Bleijenbergh and Marloes Van Engen along with their research team
used participatory modelling as an intervention methodology to "support stakeholders in:
reaching a shared problem definition and analysis of gender inequality; and identifying and
implementing policies to tackle gender inequality."

The Process: Participatory Modelling and CLD Development

Participatory modelling is a method in which different stakeholders to a problem get together
to create a shared model of the problem. This is typically done in day or half-day long
workshops, which are usually facilitated by one or two experts in systems modelling. The aim
is to increase understanding of the problem, synthesise information from different
perspectives, reach consensus between stakeholders with opposing viewpoints, and
ultimately come up with a strategy for solving the problem.

The researchers in this case study used CLDs as a modelling methodology and involved
many different stakeholders in various stages of model development. They held in-depth
interviews with 44 academics from universities that suffer from low female-to-male ratios at
the assistant and associate professor level. The results of these interviews were then
discussed with "five focus groups with 34 stakeholders."

Researchers along with stakeholders developed models to represent the problematic issue
at hand for each school independently. A generic structure was generated to represent the
common structure among these schools that gives rise to the problematic behaviour, i.e., the
slow progress towards a more equal ratio between female and male assistant and associate
professors. A set of actionable policies were then put forward by the researchers based on
insights they gained from the model, which can be seen on the next page.
The Model
A slightly simplified version of the model produced by the team of researchers is shown to
the right*. It identifies three main reinforcing feedback loops that go some way to explaining
why the ratio of female to male professors is so low in Dutch technical universities.
Note that this CLD identifies stocks and flows, which we discussed in the previous module.
In this case the participants felt it was useful to use stocks and flows to identify what is
known as an "aging chain" by which women go from being PHDs to associate professors to
full professors. Note that many outflows from one stock are an inflow to another.
We can now discuss each loop separately, which will also help us to better understand this
stock and flow structure.

*Please note: this model is our own recreation of the research team's model. It has been
altered slightly for educational purposes.

R1: Masculinity of Norms

The reinforcing loop R1 says that the low Proportion of women full professors leads to a
Masculinity of norms in Dutch technical universities. This reduces the Perceived congruence
of women with the ideal academic which reduces the number of women PHDs deciding to
become associate professors. This in turn reduces the number of Women full professors,
which further reinforces the Masculinity of norms.

By masculinity of norms, the researchers were referring to the culture of masculinity that
existed in each university. Masculine culture had different meanings to different participants,
but the researchers were happy enough to have just one variable to represent this broad
concept in this high level model.
The basis for the importance of recognising the potential masculinity of norms came from the
researchers' interviews. For example, consider the following excerpt from an interview with a
male Dean at the Technology institute.

The researcher asked:

"To what extent do you think your policy creates unintended obstacles to hiring, selecting
and promoting women?"

The Dean responded:

"Culture is the only thing that springs to mind. The culture of material engineering, even
more so the culture of maritime engineering, it’s quite no nonsense. It does have its positive
sides […] but […] um […] it’s still a masculine culture."

R2: Visibility of Women

Reinforcing loop R2 was also identified as working as a vicious cycle. The low Visibility of
women academics meant that there is less Perceived congruence of women with ideal
academic role. This reduces the number of women going for associate and full professor
roles, which further reduces the Visibility of women academics.
R3: Effectiveness of Networking for Women Academics
The researchers highlighted the importance of networking in their model. During their
interviews, the subject of the importance of networking was brought up several times. For
example, consider the below two excerpts:

"Sure, you can see all kinds of alliances appearing and, if women are on the Board, they’ll
make sure that women will get into other positions." (Assistant Professor at the Social
Sciences and
Humanities University, woman).

"Traditionally it’s a bit of a man’s world, so networking and that sort of thing is traditionally
easier [for men]" (Assistant Professor at Technology University, man).

The researchers used data such as this to inform the creation of another feedback loop that
highlighted how having a higher Proportion of women full professors increased the
Effectiveness of networking for women academics, which in turn helped increase the number
of Women full professors and so further increased the Effectiveness of networking for
women academics.

As with all reinforcing loops, this loop can work as either a virtuous or vicious cycle. This is
seen in the perspectives expressed by the interviewees quoted above. Whilst the first quote
discusses the potential for women to bring more women into senior academic roles, the
second quote points out how the existence of a "man's world" can make networking harder
for women at the moment.
R4: Masculine Culture and Motivation
Another loop identified by the researchers concerns the motivation for women to have an
academic career in technology. They identified this lack of motivation as a reinforcing loop,
as shown to the right. This loop shows how the Masculinity of norms in Dutch technical
universities reduces the Motivation of women for an academic career in technology. This
reduces the number of Women full professors, which in turn increases the Masculinity of
norms further, thereby further demotivating women to go for more senior roles.

The above describes the loop acting as a vicious cycle. However it can also act as a virtuous
cycle, when more women get promoted and so reduce the Masculinity of norms, and thereby
increase the Motivation of women to have an academic career in technology.
R5: Visibility and Motivation
Similarly, the Visibility of women academics is another way in which women can become
increasingly motivated or demotivated to have an academic career in technology. The more
visible women are in academia, the more motivated those doing PHDs will be to further their
career in this area. Similarly, the less visible they are, the less motivated other women will
be.

By looking again at the whole model, we can see how all of these reinforcing loops interact.
The researchers found that the CLD helped in visualising the interaction between these
different causal structures, and highlighted how many feedback loops working as a vicious
spiral could be made to work as virtuous cycles.
Outcomes: Achieving Consensus Around Actionable Policies
The creation of the CLD in this case was said to have helped in creating a shared
understanding of the issue at hand. The fact that it was done in a participatory way was also
very important for capturing the perspectives of different stakeholders, and ensuring that
they felt that they had been listened to.

Furthermore, the identifications of interrelated feedback processes helped stakeholders

identify different intervention policies to solve the issue. The following is a list of some of the
policies generated and implemented as a result of the overall process:

"Concerning masculinity of norms, the Social Sciences and Humanities University introduced
a correction of publication productivity targets for those taking care leave and part-timers,
and for paid research sabbaticals following maternity leave. The Technology University
introduced the paid research sabbatical following maternity leave, and relieved the
prohibition of flexible positioning in the schemes."
"Concerning visibility of women, the Social Sciences and Humanities University installed a
task force that monitors the visibility of women researchers in internal and
external communication. Since 2011, selection committees (containing at least one woman)
at the Technology University have had to put forward at least three highly qualified women in
the discipline on the shortlist of possible candidates for full professors."

Reflecting on Case Study 1

Overall, the researchers in this case believed that they achieved a lot in their research, and
that the use of CLDs helped them identify important systemic structures that are inhibiting
women's progression to more senior roles in Dutch technical Universities.

The researchers mentioned how "integrating the knowledge of researchers and stakeholders
in a causal loop diagram supported the learning of stakeholders about the issue of gender
inequality." It also "supported validation of the analysis with organisational stakeholders and,
as a consequence, ensured their commitment to the results. Both universities identified and
implemented a considerable number of gender equality interventions following the
participatory modelling process."
Case Study 2: Hewlett-Packard Market Share

The Context
Hewlett-Packard (HP) is a leading manufacturer and service provider for many consumer
products. However, managers at the firm were becoming increasingly concerned about a
decline in market share for the company's products. Although HP covered high market share
for a specific product, the company was facing stiff competition from competitors offering the
same product at lower prices, and incentivising retailers to sell their product.

Deborah Campbell, a senior consultant experienced with systems modelling, suggested

participatory modelling* as a possible intervention methodology to understand the overall
system and to create a shared understanding of the problem and its possible solutions.

*participatory modelling is the broader term for all kinds of modelling done as a group. Group
model building is one school of modelling within participatory modelling.

The Process: Model Formation and Policy Generation

Deborah and the management team decided to form a team of 15-20 people from the
business to develop a CLD as well as a formal in-depth quantitative model. The objective of
the exercise was to create a shared understanding of the underlying reasons for the loss in
market share and, with that understanding, to develop a strategic action plan to address the
problem.

The members of the team came from many different departments within HP, which is
important if a systemic understanding is to be gained. Deborah explains that the process
was not easy, involving a lot of confusion and uncertainty at different stages of the model's
development. Some of the staff, inexperienced with systems modelling, pushed for too much
detail in the model, which can often hinder the model's ability to generate high-level insights.

Working Through Confusion

However, Deborah observed that confusion and uncertainty is a natural part of model
development. Having the team go through the process of model development, including
each department’s role in detail, and then passing through that confusion to reach a new
shared understanding was a very useful process. Crucially, the fact that it was done as a
team created a sense of ownership among all participants and resulted in consensus and
deeper understanding of the whole process. Even once a high-level model was formed,
participants understood the hidden structure that lay behind it.
The Model
This is the high level qualitative model that came out of the group's process. It involves the
participation of different departments and represents the group's different mental models of
how the market share of HP's product is affected by the price they set for that product, as
well as by the overall market dynamics and consumer and retailer behaviour.

This model is too large for us to look at in detail here, but it will be interesting to look at a
small part of it:

We can start by looking at the central variable in the model, which was the Probability
purchaser buys HP supplies. In this case, the purchaser is most likely to be a retail store,
who will then sell the product to consumers.
The team argued that the Probability purchaser buys HP supplies is affected by the End user
and purchaser's satisfaction with HP supplies.

At the same time, the Probability purchaser buys HP supplies is reduced by their satisfaction
with competitor's supplies. The more satisfied the retailer is with competitor's supplies, the
less likely the retailer is to purchase supplies from HP. In essence, this is simply saying that
HP are competing for market share with other firms.
Ultimately, the probability of the purchaser buying HP supplies over their competitor's
supplies will affect HP's sales, which will affect their profits, as modelled to the right.

The team recognised that their profit will affect their investments. This is turn will affect their
ability to communicate relative value to the retailer. In other words, an increase in profit will
allow them to invest more in marketing and sales.
This in turn will increase the Retailer perception of our relative value proposition, which will
increase the Probability purchaser buys HP supplies. This creates a reinforcing loop in the
system, whereby HP's profits determine their ability to invest in marketing and sales, which
in turn affects their sales and so their profits. When working as a virtuous loop, this will
increase HP's market share exponentially. However, if competitor's get a better grip on
market share, this will hurt HP's profits and so their ability to invest in marketing and sales,
potentially sending their profits into a downward spiral.

Outcome: Consensus and Action Plan

The management team was extremely impressed with the high-level model, and used it to
develop a formal quantitative model for policy evaluation. Moreover, the model created a
common communication tool among different departments to discuss different strategies and
outcomes. The model was also helpful in bringing about many useful insights. For example,
participants in the process explained how:

‘‘The model has highlighted the complexities, interdependencies, and gaps of knowledge
about our business.’’

‘‘The factors enabling consumers to switch [to competitive supplies] are not being addressed
aggressively enough.’’

‘‘We were reminded of the importance of influencing both supply product attributes and the
consumer perception of supply product attributes simultaneously.’’
Moreover, the management found the model development journey to be useful. The journey
through confusion and uncertainty created a shared understanding among all departments
about how the overall system works, and how each department has a role in the company's
market share. Furthermore, the model showed the interrelationships and complexity of the
overall structure, emphasizing the importance of a unified action plan. Crucially, developing
the model with stakeholders created ownership and facilitated the implementation of the
resulting action plan.

Many employees also felt that the model helped them to see the bigger picture of the
business they were in. One commented about how the model ‘‘ makes you think beyond
your job’’ and another spoke of how the model helped them to ‘‘see how what you do and do
not do, affects the business’’.

Reflecting on Case Study 2

The HP case study demonstrates how CLDs can be used as a means of synthesising
information and perspectives from different departments in a business. Interdepartmental
communication is increasingly recognised as important for business success, and cases
such as these demonstrate how CLDs can serve as a valuable methodology for structuring
this communication.

To read more about this case, ​please see this article which describes the case.

Case Study 3: The Martina Bustos Community

The Context
The case of the systems modelling with the Martina Bustos community was described in one
of the final sections in Module Four. In case you missed it there, the below text describes the
context of the situation once more.

Martina Bustos is a settlement that is located 4 km from downtown Liberia, the capital of the
Guanacaste region on Costa Rica. This settlement was established about 15-20 years ago
when a private landowner, named Martina Bustos, donated the territory so that low-income
families could have a place to live. However, Martina never formally handed over the land
titles, and to this day they remain with her children.
Image courtesy of The Sustainability Laboratory
Nonetheless, many poor migrant families have moved to the area, and the population on the
settlement has grown to between 3000-3500 people, mostly of Nicaraguan descent. The
lack of land titles has caused many problems as, without them, the community is not formally
recognised by the Costa Rican government. As a result, there is no legal obligation for the
government to provide access to basic services such as electricity, waste management, and
clean water.

In the summers of 2014 and 2015, participants in the Sustainability Lab’s Global
Sustainability Fellowship program began working with the Martina Bustos community. The
purpose of the work was to create a community development plan that would serve as the
blueprint for creating a more robust and sustainable community. The fellows worked with
members of the community (mainly the women’s association) to identify the numerous
problems in the community, and try to find a strategy for resolving them. Much work had
already been done by other NGOs in this area, but The Lab’s approach was unique in that it
took a holistic/systems perspective on the community, addressing multiple problems at once
(economic, social, environmental, motivational) rather than focusing on just one issue at a
time (for example, focusing on just housing or just water, as had been the focus of some
previous work in the community).
Image courtesy of The Sustainability Laboratory

The use of CLDs

As part of this work, the fellows held Group Model Building sessions with the community
members, the purpose of which was to discover together how the problems of the
community might be interlinked, and how solutions to one issue would not suffice if
considered in isolation to other issues. For these sessions, about ten community members
were invited to the local university. They participated in a two hour session which was
facilitated in Spanish by Vanessa Armendariz, one of the fellows.
Image courtesy of The Sustainability Laboratory

Vanessa began by asking the community members which issues in the communities were
most important, which was decided on by a voting system. After a brief explanation of the
purpose of the session, and a brief introduction to CLDs, Vanessa then wrote these issues
on a white board, and began asking the community members to identify the immediate
causes and effects of these issues. As they spoke, Vanessa began drawing arrows from
causes to effects, stopping to ask participants if she was properly representing what they
were saying.

A few hours of intense discussion later, the community had developed their own CLD that
showed how the seemingly separate issues of the community were in fact heavily
interlinked. The CLD also helped to identify many vicious cycles happening in the
community, which could be turned into virtuous cycles if acted upon in the right way. New
iterations of this CLD were then used by fellows as a basis for creating a strategy of
community development in Martina Bustos.

The issues addressed in this session were quiet broad, with everything from the land rights
to education levels to environmental problems (such as waste disposal) being represented.
For this case study however, we will focus on just one part of the CLD, which represented
the dynamics around community organisation and the land rights issue.
One of the issues identified by the community was the number of people with land rights.
The land titles remained with the children of Martina Bustos, and this put the community
members in a precarious position, as they felt they could be removed from their homes at
any moment.

One effect of this was that the sense of belonging of each individual was very low.
This lack of individual belonging then decreased the overall sense of community, as people
often felt that their people were transient rather than permanent community members. As
such they felt there was little sense in getting to know and rely on each other too much.

The lack of sense of community was a big factor in the lack of community organisation. For
example, there was only one organised group in the community, known as the women’s
association, and they struggled to retain more than 5 or 6 regular attendees to their
meetings. Many residents of the community said they were either not interested enough to
attend these meeting, or were simply too busy. Moreover, many said that until they got their
land titles, there was little point in putting effort into improving the community.
However, at the same time, many community members identified that the owner of the land
titles, Martina’s son, would be much more likely to hand over the land titles if he saw a more
organised community. Indeed the son had often expressed doubt about whether handing
over the land titles to such an unorganised community was a good idea.
In identifying this, the group realised that there was a reinforcing feedback loop that was
currently working against them. They saw how the community were not likely to organise
themselves unless they got land titles, due to lack of communal sentiments. And at the same
time, they saw how the land owner was not likely to hand over the titles until the community
became more organised. Both the landowner and the community were stuck at an impasse,
and one of them would have to move first to make things work

.
In addition to getting the land titles, the group also talked about how good community
organisation can make them more likely to get government support for community projects.
Currently, such projects are not supported by the government, because they don’t believe
the community can be organised enough to see their projects through and to maintain them
when in place.

This lack of political leverage makes it less likely that there will be successful community
projects, which in turn reduces the sense of community cohesion, further exacerbating the
situation.
.
At the same time, the community pointed out that a lack of successful projects had led to a
lack of community infrastructure. For example, there were only three or four places to get
clean drinking water in the community, and the school was quite far away. Tasks like
fetching water and getting the kids from a far away school were quiet time consuming, and
this reduced the time they had for attending community meetings and helping out with
community projects, which was another block towards communal action.
In addition to a lack of successful community projects, the lack of land titles meant that the
government had no mandate to provide access to basic services such as clean water or
electricity.

With the identification of all these interlinked feedback loops working against them, the
community could see an opportunity to turn things around. They could see how more
community participation could lead to more likelihood of receiving the land titles, which could
increase the community participation of others. At the same time, they could see the
potential for political leverage, which in turn could help with community infrastructure.
Infrastructure could also help by freeing up community members’ time that would otherwise
be spent fetching water or collecting their children from a far away school.

Outcomes
One of the most important outcome of the use of CLDs in this process was the dialog that it
facilitated between community members and the fellows. Fellows were able to dig deeper
into the communities issues, and community members were able to hold a fruitful discussion
about the inter-linked nature of their issues. Simply having a facilitated conversation with the
common goal of producing a model of the community's problems seemed to allow for a more
fruitful discussion than usual, which is indeed one of the strengths of group model building.

Additionally, the CLDs played a crucial role in the formation of the strategy that was created
by and with the community. This strategy has been described in this document, which
outlines a blueprint for improving conditions in the Martina Bustos community. The Lab is
currently seeking political and financial support to help implement this strategy.
Reflection on Case Study 3
The Martina Bustos case study demonstrates how causal loop diagramming can help a
community to not only communicate more effectively between themselves, but also to
synthesise their different perspectives on an issue so as to realise new solutions. In this
case, the community realised the chick and egg situation they were in in regards to the land
titles issue. They saw how the land owner would not hand over the titles until the community
organised itself more, but they also saw how many community members were not willing to
participate in community building activities until they had the security of the land titles.
Seeing this as a reinforcing feedback loop was an insightful moment for the community
members and for the fellows of The Lab.

Case Study 4: The ISLANDS Project

Introduction
Small island states are specifically vulnerable to climate change, natural disasters, and
external shocks such as energy crises and financial volatility. Part of this is due to their small
size, lack of access to markets, financial resources, and their limited human and institutional
capacity.

This vulnerability has been specifically recognized in various environmental and

development strategies. As a result, initiatives are forming to increase the resilience and
achieve sustainability objectives in these states. One such example is the ISLANDS project,
aiming to contribute to an increased level of social, economic, and ecological sustainability,
while simultaneously strengthening cooperation, in the region of Southern and Eastern Africa
and Western Indian Ocean. More specifically, the project was carried out in Comoros,
Madagascar, Mauritius, Zanzibar, and Seychelles. Here, we will go through the process,
have a look at how system dynamics was used, and explore the results and outcomes of the
process for Seychelles. For the full story, ​see this article​, by Deenapanray and Bassi (2014),
or section 4.3.1 of ​this book​ by Gilbert Probst and Andrea Bassi.
Figure 1. Seychelles is one of the world’s smallest countries, compromising 115 islands, and
located in the Indian Ocean - east of Kenya and north east of Madagascar. The country
holds a remarkable biodiversity, where the wildlife includes sea turtles, tree frogs,
chameleons, and over 900 species of fish. Image courtesy of Encyclopaedia Britannica.
The Project Design
The ISLANDS project was designed based on three pillars. The first important underpinning
was resilience and complex systems theory. When employing this perspective, sustainability
is seen as an emerging property of complex social-ecological systems, where the views and
conceptual understanding of sustainability among the actors in the systems shape the future
development. Moreover, the project made use of a multi-stakeholder process, to establish
cooperation and a community of practice among the participating countries. Lastly, capacity
building and learning-by-doing were crucial to the project development.

The ISLANDS project consisted of four key steps. First, a monitoring system was set in
place, to evaluate the implementation of the sustainability strategy on national, regional, and
international levels. Second, best practices in mitigation and adaptation to the vulnerabilities
of the island states were identified through data collection and modelling. Policy options and
potential interventions were assessed using system dynamics modelling, through a lens of
local culture and values. Therefore, it was specifically important to acknowledge different
disciplinary perspectives, and to include both expert and non-expert knowledge in the
process. Third, a commitment to action was established, and financial support leveraged.
Lastly, the partnership between the actors engaged in the project was strengthened.

The Modelling Process

The aim of the modelling process was to build ownership and trust, shared knowledge about
the policy context in the different countries, and from there conduct policy analysis to support
small island states in their work to achieve sustainability objectives. The participants in the
model building sessions included representatives from academia, the government, civil
society, and the private sector.

In the working group in Seychelles, there was in total 15 people taking part in the process.
The CLD created by the group included the interactions between three economic sectors
identified as important in the country; the fishery sector, tourism, and the financial sector.

The Model
The figure to the right displays a simplified version of the CLD, highlighting some key
feedback loops, which we will discuss individually.
R1: Reinforcing Tourism Infrastructure
As depicted by the reinforcing feedback R1, an expansion of the tourism sector can be
reinforced by making investments to improve service provision. The larger the tourist
industry revenues, the greater the ability to invest in the quality of hotels and tourist
infrastructure. This provides an opportunity to charge a higher price for hotel stays, in turn
generating even larger revenues for the tourism sector.
This loop could also work the other way, whereby a decrease in Tourist industry revenues
could reduce the Quality of hotel infrastructure, thereby reducing the Average price per night
and further reducing the Tourist industry revenues.
B1: Price Balances Demand
However, higher hotel prices will also deter tourists, thereby lowering Tourist industry
revenues, thereby lowering the ability to make further investments in Quality of hotel
infrastructure, which forces companies to lower the Average price per night. This is shown by
the balancing feedback B2.

B2: Tourism and Environment

A main driver of tourism in Seychelles is the natural environment, offering beautiful
surroundings and a rich wildlife. However, an expanding tourist industry results in increasing
resource consumption and negative environmental impacts, thereby lowering the
attractiveness of Seychelles as a tourist destination. This in turn reduces the arrival of
tourists. This dynamic is represented by the balancing feedback B2.
Putting these insights together we can see how there is a complex interplay between the
size and quality of the tourism industry, and the state of the natural environment. In this case
the CLD was much larger and represented many more sectors of the economy, but we just
wanted to give a small preview of the kind of modelling that can be done in such cases.
Outcomes
The ISLANDS project had multiple outcomes. First, new forms of co-operation were
established, with expert groups from the participating countries now meeting regularly and
engaging in knowledge sharing. Additionally, through the modelling process data gaps were
identified, which could guide future research efforts. Lastly, steps were taken to
institutionalize the use of system dynamics modelling as a tool to support policy design and
testing, particularly within the context of national development planning for Seychelles and
Mauritius.

Figure 2. The economy in Seychelles is heavily dependent on the tourist industry,

representing around 25% of the total GDP. Image courtesy of Encyclopaedia Britannica.

Reflection on Case Study 4

The ISLANDS project demonstrates how CLDs can be used to bring various stakeholders
together to analyse a problem from a multi-disciplinary perspective. Such analyses are
increasingly recognised as vital to the sustainability movement, and the ISLANDS project
offers another example of how CLDs can be used as a tool to help foster these more holistic
understandings of complex social, economic, and environmental issues.
Case Study 5: Building a Healthcare System

Story Background
Healthcare systems are most certainly complex, and managing such a system is
challenging. The task becomes even more difficult when one is faced with a decline in
budget. This was the situation faced by Dr. John Hunter (JH). John was the lead manager
and developer of an oncological care unit (a unit for treating cancer) situated in the
department of internal medicine at the central hospital in Klagenfurt, Austria. He had
pioneered and directed the development of this oncological care system over roughly 25
years , covering the whole state and involving ten hospitals as well as multiple local
physicians.

In 2009, John learned that the budget for the oncology unit was going to be reduced, as the
administration of the central hospital argued that it needed to "improve the economic
situation."

The Methods and Frameworks Used

John had been using the help of an organizational scientist, Mark Stone (MS), since he first
took up the challenge of developing the oncological care unit. They had worked in
conjunction to conceptualize and organize the unit, and they had relied heavily upon the
method of system dynamics and the viable systems theory (mentioned in Module One).

In this case study, we will focus on the use of causal loop diagramming - a method of
qualitative system dynamics - in the project. However, if you are interested in learning about
the wider theoretical context behind the study, ​then you may want to read this paper.

The Use of Causal Loop Diagramming

John had used system dynamics modelling from the outset of this project , as a method for
visualizing and accounting for the interdependencies between the many different parts of the
healthcare system, some of which were quite tangible (e.g. available infrastructure) and
some of which referred to less tangible but equally important variables (e.g. staff motivation
and the reputation of the oncology unit).

When the budget cut was announced in 2009, John decided that system dynamics could be
a useful tool in analyzing the potential consequences. The results of this analysis could then
be presented to the administration team at the central hospital, so that they would be aware
of the implications of their decision.
The Model
Through their analysis and conversations with oncologists, JH and MS identified some
noteworthy dynamics that would likely arise as a result of the budget cut. We can now
explore this model one loop at a time.
It's clear that a lower budget would reduce the Personnel (i.e., the staff in the care unit)
available. This would then reduce the level of Experience available in the organisation, which
would cause more Stress in an already stressful environment, as the staff's capacity to deal
with problems effectively drops. The increase in Stress would then cause an increase in
Exits, as overly burdened staff looked for new jobs, or have more stress-related absences.

This structure highlights the potential for a downward spiral to occur, whereby a small initial
reduction in staff (due to the budget cut) could result in a continuous and accelerating
decline in staff, as stress levels pushes more and more people out the door.

At the same time, a reduction in Personnel would reduce the patient-doctor ratio, which
would reduce the Quality of care and thus the Success of care. This, in turn, could reduce
the number of Healings, which would slow down the outflow of patients, and thereby further
reduce the doctor-patient ratio, i.e. the Personnel per patient.

This highlights how the reduction in budget could cause another downward spiral in the
system.
Similarly, a reduction in the Quality of care would not only reduce the Success of care, but
would also increase the Duration of treatments. This would also slow down the Healings,
causing the number of patients to accumulate.
Taking a look at the whole model, we can see how loop R1 is also linked with loops R2 and
R3, as the reduction in Personnel, and the corresponding increase in Stress, will affect the
Quality of care in various ways.

Together, these loops represent the dynamic hypothesis of what the oncologists thought
would happen as a result of the budget cut: a loss of Personnel and therefore Experience
will be further reinforced due to the increasing overload and Stress and therefore Exits of the
personnel in the ontological care unit. This will lead to a decay in the Quality of care, which
will result in a growing load of patients to be treated. This not only further increases the
burden on existing staff, but also increases the hospital's costs, as patients spend longer and
longer in the hospital.

The result is that the budget cut would actually worsen the economic situation by increasing
the hospital's costs, as well as reduce its income by slowing down the rate of successful
treatments. Thus the model shows how the decision to cut the budget would have the
opposite effect than desired, due to the dynamics described above.
Outcomes
To provide a more rigorous perspective, a simulation model was built as a next step in the
process. The results of this and the qualitative CLDs were then presented to the staff of the
oncological unit, who agreed with the analysis and appreciated the insights gained from it.
However, when the model and its insights were presented to the central hospital
administration, it seemed to have little to no impact on their decision regarding the budget
cut.

The authors of the case report speculate that there could be two reasons for this. One is that
the administration teams were unreasonable and stubborn. The other, perhaps more
elaborate reason, is worth hearing in full, as it represents an important piece of wisdom that
has often been learned the hard way in system dynamics studies so far. The authors
explained that:
"people are motivated by a sense of “psychological ownership,” a feeling that they can lay
claim to certain organizational factors as their own (Pierce et al. 2001). Such ownership
fosters self-identification as well as organizational commitment and citizenship behavior (Van
Dyne and Pierce 2004). The managers who initiated the budget cut had in no way been
involved in the formation of the simulation model. That model had been developed rather
spontaneously between the two authors. They (the authors) did not consider, at that point,
what in hindsight appears as a straightforward way to proceed: trying to involve the
administrators in the construction of the model and using the model in the political
negotiation process."

This led the authors to conclude that "the improvement of a management system can be
best obtained not only by designing a dynamic, multidimensional model but also giving
decision-makers a sense of ownership of the system under development through their
participation in the design process. Good models and logically cogent argumentation, based
on careful studies at the content level, are necessary to trigger change in direction of a
desirable future. But they are not sufficient. A context needs to be created that makes these
arguments effective. In our case, the management of the hospital was not prepared
adequately. They enacted their decision without an adequate understanding of its
implications."

Reflection on Case Study 5

This is an important conclusion reached by the authors. It is now widely accepted in the
system dynamics community that including all relevant stakeholders in your model building
process is of utmost importance. This reflects system dynamics' assertion that action and
decision making are based on deeply ingrained assumptions, known as mental models.
From experience it seems clear that mental models cannot be changed based on logical
advice alone. Rather, mental models can only be changed when we go through the process
of learning ourselves, and feel a sense of ownership about the way in which an analysis was
conducted, and the outcomes it proposes.

As such, if we want our analyses to have an impact on anyone's mental model, then they
should ideally be included in the process of model building, rather than simply being shown
its results. This is a valuable lesson to learn.

Reflecting on Module Five

That brings us the end of Module Five. We hope that this module has given you a brief
oversight on the numerous ways in which CLDs have been used to good effect in the past,
and we hope that you have taken away some valuable insights from studying these cases.

There are many, many more examples out there of CLDs being used to create change in
complex systems. Below you can see links to some such cases.
This case demonstrates how the participatory modelling approach (using CLDs as the
modelling language) can be used to analyse the effectiveness and potential unintended
consequences of policies in large social organisations.

This case demonstrates how CLDs and participatory modelling have been used to identify
feedback processes underlying ecosystem services and to foster a shared understanding of
effective strategies to protect these services.

This case describes how researchers have used CLDs to explore the relationships between
ecology and global finance.
Conclusion to the Course
Congratulations, you have now reached the end of the course! For a final exercise to help
you reflect on what you may have learned, please go to the next page.

Image courtesy of Larry Chen from Vancouver, Canada, via Wikimedia Commons

Closing Thoughts and a Final Exercise

We hope that you have found this an interesting and inspiring course, and we hope that it
will lead you to look further into the fields of systems thinking and system dynamics.
Learning to think in systems and to model them is a truly iterative process. This course has
introduced the basic concepts of qualitative modelling, but our experience has been that
one's understanding of these basics concepts becomes more and more nuanced over time.
The importance of including stakeholders in an analysis, of recognising systems archetypes,
feedback loops, delays between causes and effects; these are just some of the concepts
that we have found take on a new level of meaning as they are encountered in more and
more contexts. We note how we have begun to see the world in a new way: to see feedback
loops in everything from our own psychology to the economy, to recognising the implications
of delays between causes and effects, to communicating and reflecting on our own mental
models in a new way.
In this spirit of reflection, we ask that you spend a few minutes re-visiting the answers that
you wrote in response to the list of questions presented at the start of Module Two. Do you
think this course has had any effect on the way you might approach dealing with questions
such as these? Would you now spend longer thinking about potential unintended
consequences in each situation, and would you even go so far as to build a model to help
you think through these issues?

Next Steps
As mentioned on the previous page, model building and systems thinking are skills that are
developed with practice, and we hope that this module has given you a firm grounding to
further develop those skills. There are many great resources out there that will help you
develop your skills in system dynamics and systems thinking in general. Should you be
interested in further pursuing your interest in system dynamics, you can find a list of courses
offered on this page of the system dynamics society website. Below you can also find a list
of materials that can help you further develop your skills.

Further Materials:

A course, which was also developed by Loops Consulting for The Sustainability Laboratory,
uses system dynamics simulation models to increase understanding of The Lab's definition
of sustainability and related 5 core principles of sustainability. It can serve as a good next
step to explore quantitative system dynamics, as well as new concepts in sustainability. It
can be accessed here.

This book by John Sterman serves as a great introduction to system dynamics and systems
thinking (we recommend chapters 1, 3, and 5 in particular).

This website​ offers many enlightening articles on systems thinking and modelling.

This book by Gilbert Probst and Andrea Bassi provides a step-by-step approach to using
systems thinking to solve complex problems in socio-political as well as business
environment, and is a great way to explore some of the most recent applications of systems
thinking and modelling in addressing pressing modern day problems.

This paper by Michael Ben-Eli serves to demonstrate how the systems perspective can bring
about a more rigorous understanding of sustainability.

This website highlights some projects undertaken by a consulting firm using the systems
thinking methodology. It serves as another example of how these methods can be applied in
fruitful ways to help manage complex issues at the level of governments and other large
organisations.
Acknowledgements
We would like to extend our sincere thanks to the Foundation for Advancement of
Systems-Oriented Management Education at the University of St. Gallen, for their support in
the development of this course. We would also like to thank the following individuals: Gilbert
Probst, Markus Schwaninger, Andrea Bassi, Arielle Angel and the fellows of the 2018 Global
Sustainability Fellowship program.