System dynamics
Выбери формат для чтения
Загружаем конспект в формате pdf
Это займет всего пару минут! А пока ты можешь прочитать работу в формате Word 👇
SYSTEM DYNAMICS
We are going to describe in the following pages an approach to interpret the reality.
It is likely not to exist what we might call the correct or the best way to observe the reality,
given that it is impossible to point at a certain direction as the best or more adequate.
However, this is a new concept for many people. In my opinion, it‟s a useful way to
deal with the problems and challenges we are facing at the turn of the millennium: hunger,
poverty, the degradation of the environment, war, etc. It would seem that the traditional
ways of dealing with such issues haven‟t improved matters significantly.
The viewpoint presented here has several names. Here, we will call it system
dynamics. I‟m aware that “systems” could mean various things, and I will attempt to
clarify my interpretation of the word through discussion and examples further on.
As an introduction, we will look at the characteristics of models that represent the
world as a whole; as a global system.
We will then describe the characteristics of the situation of the world today from
this viewpoint.
System Dynamics
We are increasingly aware that we live in a very complex and constantly changing
reality, and that it‟s more so every year. In order to make the decisions that are continually
asked of us, we use mental models. However, these models don‟t always bring us closer to
solving the problem, as the solution may be, as Jay Forrester calls it, counter-intuitive,
even in the simplest of cases.
For instance, during a visit to the Science Museum with our children, we may have
to explain why the hole in a water tank which is nearest the ground spouts water further.
We may also have to explain why the image in a magnifying glass is inverted after a
certain distance, instead of growing indefinitely.
As Ludwig von Bertalanffy notes: for those that wish to study science and only
science, any posterior question makes no sense. "Quod non est in formula non est in
mundo". Such is the only legitimate position of science. Despite this, if we wish to further
our understanding, there is only one analogy that can explain that which is irrelevant to the
physicist, the analogy of the only reality we know directly: the reality of our immediate
experience.
All interpretations of reality, to use Kant‟s expression, are an adventure of reason.
There are therefore only two alternatives to choose from: either we reject all interpretations
of the essence of things, or (if we do attempt an interpretation) we must remain conscious
of its analogous nature, since we don‟t have the slightest proof that true reality is the same
as that of our own internal experience.
When faced with the common occurrence of a reality with a limited number of
parameters, especially when these are quantifiable, we employ formal models which allow
us to act with a reasonably high probability of succeeding. However, when faced with
complex situations with an uncertain number of parameters that are difficult to quantify,
we can resort to less formal models that provide a more structured view of the problem, its
key aspects, and possible solutions.
Lynda M. Applegate states that computers today are designed to treat information
sequentially, instruction after instruction. This works well if the problem can be structured
and divided into a series of stages. It doesn‟t work well with complicated, unstructured
tasks which require intuition, creativity and discernment.
The main application of System Dynamics is in this kind of complex and loosely
defined environments, where a human being‟s decisions tend to be guided by logic. We
must remember that science is currently based on measurable and reproducible
phenomena. As specialists in marketing know, people also behave according to certain
rules, which are fairly easy to measure and reproduce: market law (more demand pushes
up prices, etc.).
With reference to this, Javier Aracil states in his book Introducción a la Dinámica
de Sistemas (Introduction to System Dynamics) that computer models can provide
information not attainable via mental models: they can show the dynamic consequences of
interactions between components of a given system. When assessing the consequences of
certain actions, the use of mental models means running the risk of obtaining erroneous
conclusions. Intuition isn‟t reliable when the problem is complex. One possible reason for
this is that we tend to think in terms of one-way cause-effect relationships, forgetting the
structural feedback which almost certainly exists in such a system. When preparing a
computer model, we must consider each step separately. The mental image we have of the
system must be developed and expressed in a language that can be used to program the
computer. Normally, any consistent and explicit mental image of any system can be
expressed in this way. The mental images that we have of real systems are the result of
experiences and observations. The explicit formulation of these experiences in a computer
program forces us to examine, formalise and focus our mental images, thus providing us
with a greater understanding through several perspectives.
Mathematical models, which are programmable, are explicitly expressed. The
mathematical language used to describe the model leaves no room for ambiguity. A
System Dynamics model is more explicit than a mental model, and can therefore be
expressed without ambiguity. The hypotheses upon which the model is built and the
relationships between its constituent elements are present in complete clarity and are
subject to discussion and revision. For this reason, a model‟s forecasts for the future can be
studied in a completely precise way.
It‟s important to differentiate between the following two kinds of model:
predictive models are designed to offer precise information of the future status of the
modelled system, whereas management models are basically designed to decide
whether option X is better than Y. Management models don‟t require as much
precision, since comparisons are equally useful. System Dynamics models are of the
latter type.
As explained above, I understand the word system to mean a set of independent
elements that interact with each other in a stable way. The first step towards understanding
the behaviour of a system would be to define its constituent elements, and their possible
interaction. The notion of Aristotle that the whole is more than the sum of its parts takes on
a special meaning here.
The standpoint of System Dynamics is radically different to other existing
techniques for the construction of socio-economic system models such as econometrics.
Econometric techniques, which are based on behaviourism, use empirical data such as
statistical calculus in order to determine the meaning and correlation between the various
factors involved. The model is developed from the historical evolution of variables that are
declared independent, and statistics is applied in order to determine the parameters of the
system of equations that link them to other independent variables. These techniques can
establish the behaviour of the system without the need for information regarding its
internal functioning. This is how stock market models analyse the upward and downward
trends in the values of shares, the rising and falling cycles, etc. They are designed in order
to minimise the risk of losses, etc. They don‟t attempt to gain any detailed knowledge of
the internal workings of the firms, as the value of a given company rises and falls
according to its new products, new competitors, etc.
The basic objective of System Dynamics is different. It aims to gain
understanding of the structural causes of a system‟s behaviour. This implies
increased knowledge of the role of each element of the system, in order to assess how
different actions on different parts of the system accentuate or attenuate its
behavioural tendencies.
One characteristic that sets it apart from other methods is that it doesn‟t aim to give
a detailed forecast of the future. Using the model to study the system and test different
policies, we will deepen our knowledge of the real world, assessing the consistency of our
hypotheses and the effectiveness of each policy.
Another important characteristic is its long-term perspective, meaning that the
period studied is long enough for all significant aspects of the system to evolve freely.
Only with a sufficiently broad time scale can the fundamental behaviour of a system be
observed. We mustn‟t forget that the results of certain policies are sometimes not the most
appropriate, if the time horizon of the decision-making process was too short, or if there
was a lack of perspective when the problem was addressed. In these cases, it would be
useful to know the long-term consequences of actions taken in the present, and this can be
more tangibly attained if we use a suitable model.
The long-term development will be understood only if the main causes of any
possible changes are identified. This process is facilitated if the appropriate variables are
chosen. Ideally, the limits of the system should include the whole set of mechanisms that
are responsible for any important alterations in the main system variables over a broad time
horizon.
System Dynamics allows the construction of models after a careful analysis has
been conducted of the elements of a system. This analysis allows the internal logic of the
model to be extracted. Knowledge may then be gained of the long-term evolution of the
system. It should be noted that the adjustment of the model according to historical data is
of secondary importance, the analysis of the internal logic and the structural relationships
within the model being the key issues involved in its construction.
Note: All teaching material, and that includes this text, should be objective. This
text aims to be so, but the author admits he hasn‟t always succeeded. For this he must
apologise. Readers are invited to make their own assumptions as to what is an exposure of
methodology, and what amounts to personal opinion.
Identifying the Problem
What is the problem?
We are going to learn a method for constructing simulation models that help us
determine the best solution for a given problem. These are therefore management models,
not predictive models.
Firstly, we have to identify the problem clearly, and give a precise description
of the aims of the study. It may be obvious, but it‟s very important that the definition
of the problem be correct, since all further steps depend on this. This is also very
useful when establishing the amount of time and money that will be spent creating the
model.
Once the core of the problem is defined, a description must be completed, based on
the knowledge of experts on the subject, basic documentation, etc. The result of this phase
should be a preliminary perception of the elements that have a bearing on the problem, the
h-y-p-o-t-h-e-t-i-c-a-l relationships between them, and their historical behaviour.
The historical reference of a system is a record of the historical behaviour of the
main elements that are believed to influence the problem. Where possible, they should be
quantified. This is the graphical and numerical representation of the verbal description of
the problem.
It‟s a good idea to ask ourselves whether it is necessary to construct a simulation
model in order to find an efficient solution to the problem. This is an important question.
The construction of a model is a long and costly process. It can‟t be justified if
there are other more simple ways of obtaining the same results. There are essentially two
other ways: statistics and intuition.
- Statistics, or numerical calculus methods, are very useful for solving problems
where there is an abundance of historical data, or when we can assume reality will remain
stable. For example, if you want to find out how many cars will drive past your house
today, all you need is sufficient historical data, and assuming the street hasn‟t changed,
you‟ll get a good approximation.
- Intuition has got you where you are today, so don‟t underestimate it. For many
problems, intuition provides the right answer, drawing on our experience and knowledge.
Intuition is cheap and fast. Keep using it as often as possible.
Only when we can‟t apply one of these two options with certainty must we resort to
constructing a simulation model.
Once the problem is defined, we will see that there are many directly or indirectly
related aspects, or elements, which are also interrelated. They needn‟t be clearly or
obviously interrelated. These elements constitute the system. We will now study reality as
a system.
Defining the System
What is a system?
A system is a set of interrelated elements, where any change in any element
affects the set as a whole. Only elements directly or indirectly related to the problem
form the system under study here. In order to study a system, we must know the
elements that make it up, and the relationships between them.
When we analyse a system we usually focus merely on the characteristics of its
constituent elements. However, in order to understand the functioning of a complex
system, we must focus also on the relationships that exist between the elements which
form the system.
It is impossible to understand the essence of a symphony orchestra by merely
observing the musicians and their instruments. It is the coordination that exists between
them that produces beautiful music. The human body, a forest, a country, and the
ecosystem of a coral reef are all examples of systems that are far more than the sum of
their parts.
An ancient Sufi saying can illustrate this: You can think, because you understand
“one”, and you can understand “two”, which is “one” plus “one”. However, you must
also understand “plus”. For example, in a traffic problem many related elements
converge: number of inhabitants, number of cars, the price of petrol, parking spaces,
alternative transport, etc., and it‟s often easier and more effective to attempt to solve the
relationships between the elements (“plus”) than the elements themselves.
A good method to begin defining a system is to write the main problem down in the
middle of a blank page, surrounding it with the directly related elements. The elements
which affect the main problem indirectly go around the appropriate direct elements. This
will be the system that we will study in order to consider possible solutions to a given
problem.
The Boundaries of a System
Where does the system end?
We have all heard the theory that a butterfly fluttering its wings in China could
cause a tornado in the Caribbean. In our study, however, we will include only elements
with a reasonable influence on the behaviour of a system. We mustn‟t lose sight of the
objective: to propose practical action towards effectively solving the problem at hand.
The system must contain as few elements as possible, while providing a simulation
that will truly allow us to decide which of the possible courses of action studied is the most
effective solution to the problem. The models are generally small to begin with, with few
elements. They are then expanded and perfected. Later on, elements which don‟t play a
decisive part in the problem are eliminated. During the construction of a model, there are
several extension and simplification phases, in which elements are added and subtracted.
We can‟t ignore the relationship between the consumption of petrol and lung
health. When we analyse the carbon combustion process in an electric power plant, we can
see that, apart from energy, the following is produced: ash, suspended particles, SO 2, CO2,
etc. We can also see that there is no barrier between the desired product (electricity) and
the by-products. Sometimes, the so-called side-effects are as real and as important as the
main effects. The beauty of a system in nature is that the waste produced by one process
serves to feed the next. Perhaps this is the model to follow for industrial design in the
future.
The final size of the model must be such that its main aspects can be explained in
ten minutes. Any model larger than this will fail.
The Causal Diagram
How do we represent the system?
The set of elements that bear
relation to the problem, and that
account for the observed behaviour,
along with the relationships that
exist between these constituent
elements (which often involve
feedback) form the system. The
causal diagram represents the key
elements of the system and the
relationships between them.
As discussed above, it‟s
important to draft versions that will
bring us increasingly closer to the
final complex model. The minimum
set of elements and relationships that
serves to reproduce the historical
reference of a system is that which
forms the basic structure of the
system.
Once the variables of a system and the hypothetical relationships between these
variables are known, we move on to produce a graphical representation. This diagram
shows the relationships as arrows between the variables. These arrows are marked with a
sign (+ or -) which indicates the kind of influence one variable exerts over the other. A "+"
means a change in the influencing variable will produce a change of the same direction in
the target variable. A "-" means the effect will be the opposite.
So, when an increase in A results in an increase in B, or a fall in A causes a fall
in B, this is a positive relationship, as shown below:
When an increase in A results in a fall in B, or a fall in A causes an increase in
B, this is a negative relationship, which is expressed as follows:
Feedback
What is a loop?
A closed chain of relationships is called a loop, or a feedback loop. When we
turn on the tap to fill a glass with water, the amount of water in the glass increases.
The amount of water in the glass, however, also has an effect on the speed at which it
is filled. We fill it more slowly when it is fuller. Therefore, a loop exists.
The system formed by us, the tap
and the glass is a negative loop, because it
is designed to achieve a goal (fill the glass
without spilling). Negative loops act as
stabilising elements in systems designed to
reach a given goal, like when a thermostat
in a heating system guides the temperature
towards the level specified by the user.
When we construct a model, loops
appear. For example, those formed by
ABEDA, DBED and ABECA in the
following causal diagram.
Loops are defined as „positive‟ when the number of negative relationships is
even. If the number of negative relationships is odd, the loop is „negative‟ (just as -3
multiplied by +3 gives -9).
Negative loops tend to stabilise the model, while positive loops tend to destabilise
it, independently of the basic problem at hand.
Positive loops
Negative loops
Time
Time
Real-life systems contain both types of loops, and the ultimate behaviour will
depend on the dominant type at any particular moment. When a country acquires more
arms it makes its neighbours feel threatened, and causes them too to acquire more arms.
This is a positive loop, also called a vicious circle, and grows more and more as it feeds
itself. Positive loops cause growth, evolution, and also the collapse of systems.
Naturally, socioeconomic and ecological systems are made up of hundreds of
interconnected positive and negative loops, and its ultimate behaviour isn‟t obvious.
The concept of the loop is very useful, because it enables us to start from the
structure of the system that we are analysing and work towards its dynamic behaviour. If a
system fluctuates persistently, or remains in equilibrium, or drops off rapidly, we can
identify the structural reasons and decide how to go about modifying the causal loops that
are going to influence it. This procedure can be applied to anything from the control of an
industrial process to the monitoring of diabetes or cancer, fluctuations in the price of raw
materials, or economic growth.
Yet the most important use of this concept is in understanding how the structure
of systems affects their behaviour. In the same market and in the same year, various
firms that offer the same product present very different economic results. The less
competent managers put this down to causes beyond their control – the cost of labour,
competitors, customers‟ habits, and so on – when in fact they should study why the
systems they control (their businesses) have a less competitive structure than those that
show better results.
Country A perceives that the arms race was caused by country B and vice versa.
But in reality we can also say that country A has caused its own rearmament by acquiring
arms, as this causes the rearmament of country B. Similarly, the rise in oil prices is due
both to the concentration of production in a few countries and to excessive consumption in
developed countries of a product that‟s limited, inasmuch as it isn‟t renewable.
Identifying the cause of a problem as being something that‟s not external to the
system doesn‟t tend to be very popular, as it‟s easier to blame external factors beyond our
control. The trouble is, if the exponents of the argument of the external cause really believe
what they‟re saying, they‟ll be unable to identify the true cause of the problem – inside the
system – and obtain the desired results. If the system contains the elements that cause the
problem, it also has those that can be used to solve it.
For example, a product‟s life curve can be said to be regulated initially by a positive
loop that permits rapid exponential growth, followed by a steady state dominated by a
negative loop involving the saturation of the market, and finally a usually sudden drop
caused by the appearance of fast-growing substitute products.
Lastly, note that the causal diagram is very important for explaining the final model
to the user, if he or she isn‟t familiar with this technique, as is usually the case.
The Limiting Factor
The limiting factor is the element of the system that is limiting the growth of
the system at this moment in time. There‟s only one limiting factor at any given
moment, although over a period various different elements of the system can act as
limiting factors.
Maize can‟t grow without phosphates, no matter how much nitrogen we add to the
soil. Although this fact is quite elementary, it‟s often ignored. Agronomists assume that
they know how the soil should be fertilised as they know the 20 main substances that
plants require for nutrition… but how many elements are they unaware of? Attention is
often focused on the more voluminous substances, but seldom on the truly important
parameter: the limiting factor.
In order to understand reality we have to appreciate not only that the limiting factor
is essential, but furthermore that changes also modify the elements that make up the
system. The relationship between a growing plant and the soil, or between economic
growth and the resources that sustain it, is dynamic and constantly changing. When a factor
ceases to be limiting, growth occurs and the proportion between the factors changes until
one of them becomes the limiting factor. If we can direct our attention towards the next
limiting factor we can advance towards real understanding and efficient control over the
evolution of systems.
The limiting factor is dynamic; in the growth of a plant, today it might be lack of
water, whereas tomorrow this might be solved and the limiting factor might be lack of
nutrients, and so on. There‟s never more than one limiting factor.
The Key Factors
Also called leverage points, as it‟s here that pressure or influence is exerted. A
system includes several key factors, and they tend not to vary over time. We can use
them to bring about major changes in the system with minimum effort. They can unleash
violent behaviour in the system. Each system has a number of key factors, and they‟re
neither obvious nor easy to identify.
A normal person‟s key factors will be related to their health, the family and
(hopefully) their education. They‟re the driving force behind their acts in their daily
life.
We also have to take into account that these key factors can unleash violent
behaviour. Sometimes people tolerate all kinds of humiliation, publicly and privately
alike, yet a derogatory remark about their parents can be fatal. This is, then, a key factor.
Key factors can be physical (we can stick a finger in somebody‟s ear without making them
unduly angry, but not in their eye) or psychological (a minor accident in a car can make
some people react extremely violently).
In order to attain a goal, huge efforts are sometimes made in the wrong direction.
This is especially true in the personal, social, business and ecological fields. In an attempt
to avoid this, Jay Forrester proposed a set of guidelines for the business world that can
easily be extrapolated to other areas.
1) Whatever the problem is that has arisen, it‟s necessary to know the inner
workings of the system, how it takes its decisions, how it operates. Don‟t be led astray by
indications that point towards momentary or superficial factors, however visible they may
be.
2) Often a small change in one or very few policies can solve the problem easily
and definitively.
3) The key factors tend to be ruled out, or judged to be unrelated to the problem at
hand. They‟re rarely an object of attention or discussion, and when they‟re identified
nobody can believe that they‟re related to the problem.
4) If somebody happens to have already identified a key factor, it‟s not unusual for
action to have been taken in the wrong direction, thus seriously magnifying the problem.
Models enables us to conduct sensitivity studies and see which of the system‟s
elements can have a decisive bearing on its behaviour; in other words, they enable us to
identify the key factors. However, that doesn‟t mean we can‟t advance without their help.
The peculiarity of these key factors is that they are located in unexpected points or
aspects that provoke counter-productive actions. This is difficult to illustrate with a causal
diagram. The phenomenon seems to be attributable to the difficulty of interpreting the
behaviour of a system that‟s already defined, rather than to any specific structure, as the
effect of the interrelationships is beyond our capacity for analysis (for me this means that
the system has more than four loops).
This inability to perceive and interpret the nature of the system and the identity of
its key factors makes for counter-intuitive behaviour by the system, with the result that our
actions are in the wrong direction. Let‟s take a look at some examples.
a) A car engine manufacturing firm suffered a constant loss of market share. Every
four years there was a major loss of customers who seldom came back afterwards.
According to the firm‟s analyses, the problem lay in their policy on stocks of finished
products. The company was reluctant to keep a large number of engines in stock waiting
for orders to arrive, due to the high financial cost. The policy was to keep stocks of
finished products low. This policy saved a great deal of money. But whenever there was an
upturn in the economic cycle, the firm was overwhelmed with orders that they were forced
to attend to with long delays. The customers then went to the competitors, who supplied
the engines more quickly. The firm responded to the loss of sales with a programme of
cost-cutting measures, including further reductions in stocks of finished products.
b) Dairy farms are steadily disappearing. Measures are proposed to combat this,
including tax cuts, soft loans and subsidies. There‟s plenty of incentive for anybody
wanting to start up a small farm. However, the main reason why farms close is expansion.
Farmers try to increase their income by producing more milk. When all the farmers do the
same the market is flooded with milk and prices fall (as there‟s no intervention or
guaranteed price; if there were, the burden would be shifted to the external factor). When
the prices have dropped, each farmer has to produce more milk in order to maintain his or
her earnings! Some manage to do so and others don‟t, and of the latter those that are in the
weakest position give up farming.
c) One of the key factors in any economy is the useful life of the installed capital.
The best way to encourage the sustained growth of the economy is to stretch this useful life
as long as possible. Yet the policy that‟s practised is one of accelerated obsolescence, or
priority is given to replacing existing equipment with machinery designed to provide shortterm economic growth.
d) The right way to revitalise the economy of a city and ease the problem of
depressed areas occupied by people without economic resources isn‟t to build more
subsidised housing. The solution is to demolish the abandoned factories and houses, and
create space to set up new businesses, thus allowing the balance between jobs and
population to restore itself.
Ideally, we‟d have a set of simple rules to find the key factors and know which
direction to act in. It‟s not always possible to find these points by simply observing the
system, and this is where computer simulation models really come into their own.
Classification of Systems
Stable and Unstable Systems
A system is stable when it consists of or is dominated by a negative loop, and is
unstable when the loop is positive. That is, when the dominant loop contains an odd
number of negative relationships, we have a negative loop, and the system will be stable.
The basic structure of stable systems is as follows:
Here we can see that the
system has a “desired state”
and a “real state”; these two
states
are
compared
(“difference”), and on the
basis of this value the system
takes “action” to move the
“real state” towards the
“desired state”.
In this case the initial
parameters are of relatively little importance, since the system will act according to the
environmental conditions it encounters, so if it‟s hungry it‟ll look for food, and when it
finds it, it‟ll deal with its next objective, and so on.
It‟s important to note that in stable systems the structure that generates the
behaviour is always the same: there‟s an odd number of negative relationships, and
the loop is negative.
This means that the system permanently compares its real state with the desired
state, and when there‟s a difference, it takes action to bring its real state closer to the
desired one. Once these two states coincide, any change in the real state will result in
action (proportional to the difference) to regain the desired state.
This is how we usually find systems. By the time we get close to them, they are in a
position of stability. If a system is unstable we‟re unlikely to be able to study it, as it will
have disintegrated before we can analyse it. However, if we‟re designing a totally new
system, we should take the trouble to find out whether it‟s going to be a stable one. And if
we‟re designing a change in a stable system, we have to ensure that we‟re not changing it
into an unstable one.
Examples of systems that are not in an optimum situation but carry on over the
years – i.e., stable systems – can be found in many fields: government, workers and bosses
together produce the inflation that‟s harmful to all. Rich countries and poor countries trade
with raw materials, each with a different political and economic objective, and the result is
permanent price instability.
Let‟s suppose that the government intervenes in the system with a particular policy
that puts the state of the system where it wants it. This will cause major discrepancies
between the other elements of the system, which will intensify their efforts until, if they
succeed, the system is back very close to the initial position, after each element has made a
huge effort. For example, think of the work that‟s gone into improving the traffic in
Barcelona over the last 10 years; the traffic improved for a few years after the opening of
the Ring Roads, but now we‟re faced with the same problems as before – except that they
affect many more cars.
The most effective way of combating the natural resistance of the system is to
persuade each element to change its objectives, in the direction in which we want to lead
the system. Then the efforts of all the elements will be directed towards the same goal and
the effort will be minimum for all, as they won‟t have to resist the tide going the other way.
When this can be achieved the results are spectacular. The commonest examples of this are
the mobilisation of the economy in wartime and the recovery after wars or natural
disasters.
A less warlike example can be found in the birth rate policy in Sweden in the
1930s, when the birth rate fell below the rate of natural replacement. The government
made a careful assessment of its objectives and those of the population, and found that an
agreement could be reached on the basis of the principle that the important thing isn‟t the
size of the population but its quality. Every child should be wanted and loved, preferably in
a strong, stable family, and have access to excellent education and health care. The
Swedish government and citizens agreed on this philosophy. The policies that were
introduced included contraceptives and abortion, education on sex and the family,
unhindered divorce, free gynaecological care, aid for families with children in the form of
toys, clothes, etc., rather than cash, and increased spending on education and health. Some
of these policies seemed strange in a country with such a low birth rate, yet they were
introduced, and since then the birth rate has risen, fallen and risen again.
Some systems lack feedback, and the models we build must show the fact. For
instance, if we know the initial parameters of a clam (type, weight, etc.) and we control the
environmental conditions in which it will live, we can safely predict its weight after 6
months. There‟s a “transfer function” between the start and end values, and we have to find
it, but that‟s all.
Other examples: God is someone who gets his real state to coincide with his desired
state instantly. Suicide is the response of those who perceive that they will never get their
real state to coincide with their desired state, and that therefore all action is pointless.
Please note: The more intelligent a system is (i.e., the clearer its vision of its objectives),
the more stable it will be. This is applicable to people.
Hyperstable Systems
When a system consists of several negative loops, any action taken to modify one
of its elements is offset not only by the loop in which that element is located but also by
the whole set of negative loops, which act to support it, thus superstabilising the system.
An analysis of the system can be helpful. Any complex system, whether social or
ecological, is made up of hundreds of elements. Each element is only linked to a limited
number of variables that are important to it, and which it permanently compares with its
objectives. If there‟s a discrepancy between the state of these variables and its objectives,
the element acts in a particular way to modify the system. The greater the discrepancy, the
more energetic the action taken by the element on the system. The combined action of all
the elements that attempt to fit the system to their objectives leads the system to a position
that none of the elements actually wants, but in which all of them find the smallest gap
between the parameters that are meaningful for them on the one hand and their objectives
on the other.
Why do many problems persist despite continual efforts to solve them?
As we‟ve just seen, systems base their stability on the actions of all its elements in
pursuit of different objectives, trying to get the rest of the system as close as possible to its
desired position. From this moment on, if an element of the system or an external agent
attempt to modify its stability, the other elements will take action to go back to the initial
situation, thus neutralising the action that altered its stability.
So the answer‟s simple: systems resist any change we try to introduce because its
present configuration is the result of many previous attempts like ours (unsuccessful ones,
otherwise the system would be different today) and an internal structure that renders it
stable and capable of neutralising changes in its surroundings, such as the one we made
with our action. The system achieves this as a whole, by rapidly adjusting the internal
relationships between its elements in such a way that each continues to pursue its own
goal, and together they neutralise the action exerted on them from outside.
Oscillating Systems
We will see later, in the case studies, that for a system to display oscillating
behaviour it has to have at least two stocks, which are elements of the system that produce
accumulations.
Sigmoidal Systems
These are systems containing a positive loop that acts as the dominant feature at the
beginning, causing the system to undergo an exponential take-off. Subsequently, control of
the system is taken over by a negative loop that cancels out the affects of the earlier
positive one and provides the system with stability, setting it to a particular value
asymptotically.
It‟s important to keep sight of the
fact that in this case we‟re dealing with the
same system all the time, dominated by one
part of it in one period, and by a different
part later on. So in order to regulate its
behaviour, we‟ll have to find a way to play
up or down the part of the system we‟re
interested in. We also have to be aware that
in the mid-term the negative loop will
stabilise the system at its target value. All
we can do is regulate the time scale and the
way in which the system reaches its
objective.
Generic Structures
In complex systems, we can observe the same structure: desired state - real state difference - action, over and over again in very different contexts. On top of this base
structure, generic structures have been identified that tend to appear regardless of the
object of study.
There‟s always the same “intelligent” structure that seeks to bring the real
state closer to the desired state.
Resistance to Change
When new managers joins a firm, usually with new objectives, they often find that
its employees put up resistance to everything they propose: “they already tried that, that
won‟t work here, our customers like it the way it‟s always been, that proposal is very
risky...” In short, the company acts as a system that has managed to survive innumerable
economic crises in the past, and as a structure is capable of neutralising any change,
whether from inside or outside, due to the multiple relationships between its members.
Each pursues a different objective, yet as a whole they have succeeded in endowing the
firm with stability, although that doesn‟t mean its position is necessarily the most efficient.
For this reason it‟s often wise for new managers to seek the commitment of the general
manager for their new objectives as a way of achieving a certain amount of strength and
aligning the other elements in the company towards these objectives.
Many systems are not only resistant to new policies designed to improve their state
(greater productivity, lower costs, etc.) but also show a persistent tendency to worsen,
despite the efforts to improve the situation. Examples abound in the business world:
productivity, market share, quality of service, etc. And on a personal level, we all know
somebody with a tendency towards obesity in spite of repeated diets.
Erosion of Objectives
The action required to shift
the real state towards the desired
state always demands an effort.
And this effort in turn requires a
consumption of time, energy,
money, etc.
It‟s normal for the real
state to “contaminate” the desired
state, that is, for the system to try
and avoid the consumption of
energy required to take the action. The desired state is initially reconsidered, since if it
coincided with the real state no action would be necessary. The diagram below shows this
“contamination”.
If contamination occurs, the desired state is modified until it‟s the same as the real
state. The difference is then zero and therefore there‟s no action to be taken. And so the
real state of the system doesn‟t change.
There are only two ways of avoiding this process:
1.- Find a “hero” system. That is, convince the system that it doesn‟t matter how much
effort is required to reach the desired state, it just has to be reached. (Personally, I can
assure the student that this way of avoiding the contamination process doesn‟t tend to get
results in the 21st century.)
2.- Get an “external element” to serve as a reference or anchor for the desired state, so that
it can‟t be altered by pressure
exerted on the system, and so
that the system has no capacity
to alter the “external element”.
In Spain, when secondary
school students consider the
possibility of carrying on
studying at a public university,
they already have a fairly
accurate idea of the minimum
mark they have to get in their
schoolwork and the entrance
exams. Their desired state is that minimum entrance mark. It‟s not negotiable. Their real
state tends to be a lower mark than the entrance mark in the early years of secondary
education, so they perceive a difference, which leads them to take action (studying harder)
in order to get their real state to match the desired state. If students know what it is they
want to study, their family don‟t need to push them at all. The system isn‟t contaminated
because the desired state (the minimum entrance mark) isn‟t alterable.
Later, when they start at university, if you ask they‟ll say they want to be a great
professional, and that they‟re going to get an average mark of 10 in their degree. With the
first exams come the first fails, which make them: 1) study harder than they‟d anticipated,
and 2) tone down (contaminate) their desired state, from the desired 10 to the nonnegotiable minimum of 5.
The structure that brings about this behaviour is based on the idea that the system
includes a particular objective (e.g., desired weight) that is compared with reality (real
weight), and the discrepancy between these two values triggers an action, which is
proportional to the size of the gap. This is the usual pattern, seen up to now as a negative
loop that tends to gradually pull the system towards its objective if it encounters some
discrepancy. However, sometimes the state of the system can condition or modify the
desired state; either because the real state is very long-lasting, or because the action taken
involves a great effort, or indeed for some other reason, the initial goal shifts towards the
real state of the system.
This relieves the need to take action, as the discrepancy has been reduced, not
because the system has approached the objective, but because the objective has approached
the real state. As a result, the action taken is smaller.
In the case of the weight of obese people, this occurs when they accept that the
target weight was too ambitious, and that a more realistic target (a higher weight) is better.
This argument serves as an excuse to follow a less strict diet. When they see their weight
doesn‟t fall, they reconsider the ultimate target once again… and so on until they think that
actually their real weight is best, at which point they don‟t have to follow any sort of diet
(this would have involved a sacrifice).
There are plenty of examples of this pathology in environmental pollution, law and
order, traffic accidents, etc. In all of them, a poor performance becomes the standard in the
face of the effort required to do something effective.
A system that bases its objectives on reality and intends no more than to improve
on it is permanently drawn towards poor results. A system that gets its targets from outside
itself is immune to this type of process.
It may seem paradoxical, but if a student is convinced that he must pass all his
subjects in July because his father has imposed it as an immovable objective, for whatever
family reasons, it‟ll be easier for him than if he himself had made that decision. If it‟s a
personal decision it can be reconsidered when some of the subjects prove to be too
difficult. He can accept to leave one or two for next time, which means less studying.
However, if the objective is non-negotiable, this risk doesn‟t arise, and he has to study as
hard as necessary to reach the objective.
Economics provides any number of examples. In Spain nobody remembers such
low rates of inflation as we have now. Any government would be satisfied, and would be
happy to give up on reducing inflation further, as that would mean taking very unpopular
measures (a wage freeze for civil servants). If the target for inflation were in the hands of
the government, corrective measures would have been less strict in the past and the
present, since they would have meant less public spending and therefore lost votes.
However, the target for inflation was imposed as a condition for entering the euro zone and
as such was beyond the control of the government, who pulled out all the stops and took all
the unpopular measures they deemed necessary, because there was a fixed goal with a
deadline, and it was non-negotiable.
The obvious antidote to this pathology is to fix absolute objectives for the system,
that aren‟t based either on the past or the present situation, and take corrective measures
depending on the difference.
An absolute objective loses credibility if it is raised or lowered, and it won‟t get it
back. We see this sometimes when an objective is raised because the initial objective has
been reached; when this happens everybody expects the initial objective to be changed
again (but this time downwards) when the results are lower than the initial objective.
Addiction
Sometimes the real state of the system matches the desired state not as the result of
action but due to support from outside the system. This support may or may not be
permanent, and may or may not be disinterested, but the net effect is to bring the real state
into line with the desired state, resulting in zero difference, and therefore action by the
system is now unnecessary.
This phenomenon occurs when there‟s an objective that serves as a point of
comparison with the state of the system. On the basis of the discrepancy observed,
corrective measures are taken proportionally, but in this case the action taken doesn‟t serve
to bring the system‟s real state closer to its desired one but rather to create the perception
that the real system is close to the desired one, whereas in fact this action has no such
effect.
The lack of clear perception of the real system leads to a situation in which the
necessary corrective measures aren‟t taken, because the state of the system is perceived as
being closer to the objective than it really is.
When the immediate or short-term effect of the action disappears, the problem (i.e.,
the discrepancy between the real state and the desired one) reappears, often with greater
intensity, so the system reapplies some measure that appears to solve the problem
whenever the effect of the previous measure starts to fade.
Alcohol, nicotine and caffeine are obvious examples of addictive substances.
Another case that springs to mind is the use of pesticides, which eliminate, together with
the pest in question, the natural control mechanisms. As a result, the pest will reappear as
soon as the effect of the pesticide abates, but this time without any natural control.
In cases of addictive systems it‟s difficult to find suitable policies, since the action
taken offers apparent results in the short run, but once the process is rolling it‟s difficult to
stop. Obviously, the best approach is to be aware of these types of processes, in other
words, to be wary of using measures that attack the symptoms but make the system worse
when they are relaxed. Once the addictive process has been started up you have to expect
at least short-term difficulties if you plan to stop this process, be it physical pain for
somebody who takes an addictive drug, rising petrol prices on inclusion of the associated
environmental costs, or more pests and lower-quality food until such time as natural
predators return.
Sometimes it‟s advisable to wean yourself off an addiction gradually. But it‟s
always less costly to avoid the addictive process in the first place than to stop it later.
Shifting the Burden to the External Factor
As they get older and spend more and more time reading, some people gradually
get poorer eyesight. In the end they can‟t read what‟s written on a blackboard, and can‟t
renew their driving licence. So they get glasses or contact lenses. Then in one year their
eyesight worsens as much as it did in the previous 30 years. So their glasses become a
necessity not only to see at a distance but also to read a document of any sort. Apparently
this happens because for years the muscles around their eyes have been straining to
compensate their poor vision, and when this effort is no longer necessary they cease to act
and end up losing this ability totally. And before long they need stronger lenses.
This is a classic example of shifting the burden to the external factor. In this sort of
system an external force keeps the system in the desired situation. A well-intentioned,
benevolent and very effective force decides to help us to get the system where we want it.
This new mechanism works very well.
But with this process, through the active destruction of the impediments that
redirected the system towards the desired position, or simply through atrophy, the original
forces that worked to correct the position of the system are weakened. When the system
moves away from the desired position the external factor makes an extra effort, which
weakens the original forces still further. In the end the original system adopts a position of
total dependence on the external factor, as its original corrective forces have disappeared
completely and in most cases irreversibly.
It‟s easy and fun to find other examples of shifting the burden to the external factor.
Here‟s the start of a possible list.
Seeking the aid of an external factor to get the system where we want it to be isn‟t
in itself a bad thing. Usually it‟s beneficial, and enables the system to tackle better
objectives. Yet the dynamics of the system can be problematical, for two reasons. Firstly,
the external factor that intervenes doesn‟t tend to perceive the consequences of its help on
the elements of the system, particularly on those that performed the same task as itself.
Secondly, the community that‟s helped today doesn‟t stop to think that this help is
temporary; they lose their long-term perspective and so become more vulnerable and
dependent on the external factor.
The withdrawal of aid from a system that‟s being helped, whether it‟s the human
body, a particular area of ecological value or a human community, doesn‟t tend to be easy
and is often simply impossible. This process of withdrawing help without harming the
system must be based on identifying the internal elements of the system that in their
original state took care of correcting the problem, strengthening these mechanisms and, as
they begin to do their job, gradually withdrawing the help.
Short and Long-Term eEfects
A rational analysis of the problem at hand based on our capacity for synthesis and
our ability to imagine things seems to be a bad guide to find the key factors. We generally
pay attention to the components of the system and their behaviour in the short term, all on
the basis of incomplete information. Consequently, firms reduce their stocks of finished
products when sales are seen to slump, the government extends its tax reductions for small
farmers, and policies are introduced to encourage firms to replace their machinery instead
of maintaining it properly. They‟re all very reasonable policies. But there‟s still something
inside us that just might make us realise that our customers‟ dissatisfaction with our long
delivery schedules, or farmers‟ permanent concern with increasing their output, or the idea
of replacing a machine that‟s productive… all means something, but we haven‟t given it
the right interpretation.
Finally, I‟d like to say that in my opinion we have the capacity to understand not
only simple systems but complex ones too, and to find the key factors. What we don‟t
appear to have is the capacity to
articulate the arguments to convince
others or even ourselves that what we‟re
perceiving is right. We expect the
solution to be closely related to the
symptom; we expect long-term profit to
start with short-term profit, or a strategy
that‟s satisfactory for all the agents
involved. Yet we know complex
systems don‟t behave that way. So
something inside us still insists
somehow that maybe that simple,
effective solution isn‟t the best. And
then we carry on proposing policies that
can‟t work, denying ourselves other
simpler and more effective ones that
could.
We try to compete instead of cooperating, and we try to reach the limits of the
environment‟s capacity instead of admitting that we‟ve already gone too far. The results
are famine, war, pollution and depression. And right in front of us, within reach of our
capacity for understanding, stand balance between countries, peace, equality and
sustainable development.
Control Questionnaire
After reading this paragraph, it is advisable that the reader answer the following
questionnaire to look at what is understood and what must be reviewed before continuing.
a. Give some examples of SYSTEM. Remember the definition of system as a set of
interrelated elements such that one element affects the behaviour of the whole set. For
example: a city.
a.1. Name some ELEMENTS OF THE SYSTEM. For example: persons, cars, pollution,
streets, etc. Incorrect elements would be: the government, the city, Barcelona, colour,
asphalt, etc. They are valid as elements of the system if we can notice when the element
increases or decreases, improves or worsens, etc.
a.2. Name the UNITS OF THE ELEMENTS. For example: Persons: number of persons,
Pollution: nº of particles in suspension/m3, Streets: m2.
b. Give a system example that has JUST ONE GOAL, indicating the goal. For example: a
mower; goal: to cut grass.
c. Give a system example that has SEVERAL GOALS. For example: a company, where
the businessperson has the following objectives: the most profit, increase the number of
clients and increase product quality.
d. Give a system example that has GOAL EROSION. Indicate some ways to avoid
contamination (in other words, erosion) of the goal by the real situation by securing it to an
external element. For example: students usually have an erosion of their initial goal of
getting excellent grades. In this case, an external element that can prevent this erosion is
the grades of a „rival‟ student.
e. Give a system example that shows RESISTENCE TO CHANGE. For example: We
prefer to wear our old shoes because they are more comfortable than new ones.
f. Give a system example that shows ADDICTION to an external aid. Indicate how you
believe you should adjust your activity, so that you are not completely dependent on this
addiction. For example: a smoker.
g. Give a system example with a goal, a corrective action that brings that goal nearer and
some LONG-TERM EFFECTS that are the result of the action that have the opposite
effect of those observed in the short-term. For example: a person wants to have an
imposing physical appearance (goal) and takes steroids to increase muscle mass (action).
The long-term effects are coronary disease that makes it necessary to be bedridden for a
long time resulting in the disappearance of the muscle mass.
h. Give a system example indicating the LIMITING ELEMENT that prevents an action.
For example: fire does not spread because there is no more wood left; the youngster does
not study because there is no more paper; the car will stop when it runs out of petrol.
i. Give system examples and some of its KEY ELEMENTS. For example: the amount of
salt in food is a key factor for it to be edible since, if we put too much salt, no one will be
able to eat the food.