Simulation & System Modelling

System Dynamics
Feedback and Causal Loop Diagrams

HANNAN MIA
Hannan@[Link]

1
 Recommended Reading
• Mia AH (2020) Simulation and System
Modelling, 1st Edition, Istehad Inc,
Canada. Chapter
• Sterman JD (2000) “Business Dynamics:
Systems Thinking and Modelling for a
Complex World”. Mc-Graw Hill, Chapters 1
and 5
• Morecroft J (2007) “Strategic Modelling
and Business Dynamics”. Wiley Chapter 2
2
 Policy Resistance
• Organisations (Management) and / people
make decisions to solve problems and
improve situations.
• Although decisions are taken after careful
analysis and with good intentions, their
consequences are not always the expected
ones.
• In many cases, the consequences of
decisions are counterintuitive. They
worsen situations instead of improving 3
them
 Examples of Policy Resistance
• A company redesigned its production system to increase
production. Unfortunately the distribution network could
not cope with the increased production flow. So the
company spent money on ‘improvements’ that had no
real effect on its financial performance.
• Project managers tend to recruit extra staff to speed up
late projects. The new members of staff cause
disruption to the project work during their training period
and need some time before integrating the project and
becoming fully productive. The result is more delay in
the project.

4
 Sources of Policy Resistance
• The main reason for policy resistance is the
inadequacy of the mental models used to
make decisions.
• Mental models refer to our view of the world
and its structure.
• Mental models tend to be simple and cannot
cope with the complexity of the real world.
• Generally, our perception of problematic
situations is not accurate. It does not include all
the elements involved in the situation and the
structure of the relationships among these
elements.
5
 Systems Thinking
• The remedy to this problem is to perceive the
world as a set of interconnected elements.
• This means, we have to conceptualise the world
as a “System” in which elements are tightly
interconnected
• That is, an action in one part of the system will
have reaching consequences in other parts of
the system.
• This way of analysis is called “Systems
Thinking”. It means that the world is a
“complex system” in which everything is
connected to everything”.
6
 Dynamic Complexity
• Policy resistance is most important in systems involving
high levels of dynamic complexity.
• Dynamic complexity indicates the counterintuitive
behaviour arising from the interaction among a system’s
elements over time. It is present if
– the short term consequences of an action are different from its
long term consequences.
– the consequences of an action in a part of the system are
different from its consequences in another part of the system.
• Decision making performance is very poor in the
presence of dynamic complexity.
• System Dynamics (SD) is the discipline which deals
with dynamic complexity and policy resistance.
7
 The Modelling Process in System
Dynamics (1)
• Problem articulation (Boundary Selection)
– Theme selection
– Time horizon
– Dynamic problem definition
• Formulation of dynamic hypothesis
– Initial hypothesis definition
– Endogenous focus
– Mapping
• Formulation of a simulation model
– Specification of structure
– Estimation of parameters
– Test for consistency
8
 The Modelling Process in System
Dynamics (2)
• Testing
– Comparison to reference modes
– Robustness under extreme conditions
– Sensitivity
• Policy design and evaluation
– Scenario specification
– Policy design
– What if scenarios
– Sensitivity analysis
– Interaction of policies
9
 The Concept of Feedback
• Feedback is the most important cause of
Dynamic complexity. This concept of is
introduced through an example.
• Assume that a country A develops a new
weapon system. This action is seen as a threat
by country B which react by developing a new
weapon system.
Country A develops weapon system

Threat perceived by Country B

Country B develops weapon system

Country B develops weapon system

Threat perceived by Country A

Country A develops weapon system

12
 Feedback Loop
• The previous structure is called a Feedback
Loop.
• A Feedback loop represent the self reinforcing
(or self balancing) consequences of actions.
• In the previous example, the action to develop a
weapon system by country A leads this country
to develop more weapon systems over time.
• In general, each country blames the other one
for this situation although they both contribute to
the situation.
• This structure is a simple and powerful
explanation of the arm race situation. 13
Representation of Feedback Loops
• To represent a feedback loop, it is necessary to
know the nature of the relationship between
each cause and effect variable in the loop.
• If the polarity has a plus (+) sign, the cause and
effect variables vary in the same direction.
• If the polarity has a negative (-) sign, the cause
and effect variables vary in opposite directions.
• There are two basic types of feedback loops:
Positive (Reinforcing) loop and Negative
(Balancing) loop.

14
 Determining Loop Polarity
• Loop polarity is determined by the polarity of the all the
links in the loop.
• There are two ways to determine loop polarity: the fast
way and the right way.
• The fast way is based on counting the number of negative
links in the loop. If the number of negative links is
even, the loop is positive. If the number is odd, the
loop is negative.
• The right way is based on making a change to a variable
and go around the loop to determine the consequences of
this change.
• If the change is reinforced by the feedback effects in
the loop, it is positive. If the change is opposed, the
loop is negative.
15
 Positive (Reinforcing) Loop
• This loop reinforces (amplifies) any perturbation in any
variable in the loop. This reinforcing process is valid if
the perturbation consists of increasing or decreasing the
level of the variable.
• Example

16
Behaviour of Reinforcing Loops
• The behaviour is exponential growth (or decline)

Time
17
 Negative (Balancing) Loop
• This loop balances (counteract) any perturbation in any
variable in the loop. This balancing process is valid if the
perturbation consists of increasing or decreasing the
level of the variable.
• Example

18
 Behaviour of Balancing Loops
• The behaviour is goal seeking

Time
19
 Causal Loop Diagrams
• The structures representing feedback loops are
called Causal Loop Diagrams (CLDs).
• Construction of CLDs is the first step in SD
methodology.
• There are only two types of loops in any CLD:
reinforcing and balancing loops.
• Although it is easy to infer the behaviour of a
single reinforcing or a single balancing loop, it is
impossible to predict the behaviour of many
interacting reinforcing and balancing loops.
20
 Guidance for Causal Loop Diagrams
• Causation versus correlation
– The links between variables in the CLD should
represent (what you believe) causal
relationships not correlations.
– A model should represent the reality and be
able to respond to new policies not just the
past behaviour, which is determined by the
correlation.
– Moreover, correlation indicates linear
relationship between variables. This does
indicate the existence of causal relationship.
21
 Example
• There is a strong correlation between the number of
infections with tuberculosis and the demand on soda
drinks. However, soda demand does not cause new
tuberculosis infections.
• Therefore, the following link is wrong

22
 Example (Continued)
• The causality which led to the correlation is due to the
cold weather, which has an impact on both the soda
demand and tuberculosis infections.
• Therefore the correct cause effect links, which should be
included in the CLD are as follow

23
 Why Causal loop diagram?
i. Quickly capturing of hypotheses about the causes of
dynamics in a given phenomenon
ii. Represents the interdependencies and feedback processes
of the variables of a system.
iii. Very powerful and useful to capture the mental model of
stake holders including self in order to effectively start
modeling a project.
iv. Communicate the results of a completed modeling effort
v. Eliciting and capturing the mental models of individuals
or teams
vi. Communicating the important feedbacks one believes that
he/she is responsible for the problem.

24
 Labelling Link Polarity
• Labelling link polarity means using “B” if loop is
negative or balancing and using “R” if the loop
is positive or reinforcing.
• When building a CLD, make sure that the
polarity of all cause effect links is labelled on the
diagram.
• This is crucial to determine whether the loop is
positive (reinforcing R) or negative (balancing
B).

25
 Why is Labelling Important?
• This is important for two reasons
– Enable you to verify your
assumptions about the polarity of a
link
– Determine if some variables are
missing.
– Enable you to determine the polarity
of each loop in the CLD if B or R.

26
 Naming the variables
• Always use nouns when choosing a
variable name and avoid verb.
• Never use action phrases to name a
variable because in the loop’s arrows
convey the action
• Variables representing quantities should
be used
• Choose the more positive sense of a
variable (Growth not decline)
• Conceive of the possible unintended
consequences of a variable (Ad-sales,
27
cost, revenue etc.
 Naming the variables
• The variables in a balancing loop play
dynamics towards a balancing i.e. goal
seeking process. In such case try to make
the goals explicit help driving the loop.
Target employee

28
 General Considerations
• Links should have unambiguous polarities. If a link can
be sometimes negative and sometimes positive, you
have to include both links explicitly in the CLD.
• It is recommended to name the loops in the CLD. This is
important if the number of loops is important so that you
can capture the role of each loop in the CLD.
• Variables should be nouns or noun phrases and must
have a clear sense of direction.
• Represent important delays in the CLD.
• Choose the right level of aggregation so that it is easy to
understand the polarities of the links in the CLD.
• Make the goals of negative loops explicit.
• Distinguish between perceived and real conditions.
29