CS 752/IE 604:

System Dynamics
Jayendran V. (IEOR), Om P Damani (CSE)

1
 What is the Course About?

● To analyze development situations and interventions

● To learn the art and science of modeling
● Not about developing new technologies
● Training in the Art & Science of Systems Thinking
➔ Science is what we understand well enough to
explain to a computer. Art is everything else we do.
- Donald Knuth
➔ Art of Modeling + Science of Simulation
➔ What is Development? What is a System?
 Development

● Development: hard to define like ‘respect’, ‘love’, and ‘hatred’ – you know
when you see one – Positive change in the lives of people
● Focus on Analyzing Development – figure out how to answer
○ Are EVs sustainable – can they ever be sustainable
○ What is the maximum population supportable in our country
○ Why is 100% vaccination not needed for eradicating any virus
○ How are prices of anything determined: what is the invisible hand of
market
○ Why providing cheap housing without providing jobs may work against
the poor
○ What is the long-term impact of AI/ML
● Lot of development mediated through Business and Economics
 Event-Based vs. System-Based Thinking

Event Based Thinking:

● Problems as events, Solution as Quick-Fix reaction

● Belief: problems happen randomly due to uncontrollable triggers
● Belief: Quick action can save the situation
● Ignoring the system
 Event Based Thinking: Pervasive

● News of flood comes Demand river-interlinking

● Electricity becomes even more

Promote solar pumps
irregular during irrigation

● Corona cases rises Shut down trains

● Traffic congestion increases Build more roads

● Observe that most IIT students Increase minimum required

credits
take minimum required credits

Not claiming that all of the above reactions are necessarily bad
Popular World View
 Popular View of Traffic Congestion

Problem as Event Quick Fix Solution

Road Outside IIT has been widened twice before the current Metro Phase
System-level View of Traffic Congestion
District Development Plan’s View of Irrigation
A System’s View of Irrigation
Promoting farm ponds in MH
A Feedback World View : A Systems Model
Systems worldview recognises that systems are:
● Complex: inter-connected components
● Interconnection results in feedback
● Feedback gets masked by delays
● All of this happens while being subject to conservation laws
● System Thinking
○ Extends analysis boundary beyond here and now
○ Focuses on interactions/feedback between components
● System Dynamics
○ Evolution of system behaviour over time
○ Take Delays and Conservation Laws into account
System: Interconnected components; Dynamics: Resulting behaviour over time
 System: Goal-Seeking Behavior

● Each system has actors with certain goals

● Actors work hard to attain their goal when the system tries to
change
○ private players, farmers -> maximise profit
● That’s why it is hard to change Social Systems
● Explains why most policies fail, or worse, have
unintended/counterintuitive consequences
○ corruption is not the only thing to be blamed
 Static vs. Dynamic Analysis

● [Link]
[Link]
● Assumes that resources are static and current distribution has
no impact on future distribution
● Ignores "Money begets money“: system structure helps increase
inequality with time - as rich gets access to better education,
health, ability to manipulate system etc.
● Approaching an uncle for internship/job
Problems of Today are Solutions of Yesterday
Summary - Development is hard, because
Social Systems are:

● Due to bounded rationality, human intuition about working of

social systems and their long term dynamics is often wrong
● Analytical solutions to formal mathematical specifications of
social systems are often intractable
● Use of Computer Modeling and Simulation can be a powerful
tool in deciphering the complexity of social systems
 System Thinking Checklist:
Given an event/proposal

● Identify System Structure: Goal seeking behaviour

○ root causes
○ motivation of all agents
● Feedback loops: virtuous and vicious cycles
○ possible effects of the proposed actions
○ possible effects of the possible effects ..chain-reaction
spanning space and time ... transitive closure
● Conservation laws: Things do not simply disappear
○ where do they come from, where do they go
○ things stay in place unless moved out or transformed
● Side-effects: no such thing
○ there are effects that we thought of, and those that we did not
think of
 Understanding Systems using “Model”

A “model” is a:
- An abstract representation of a System (it is not “reality”)
- May be Physical, Mathematical, or Conceptual; Dynamic or Static
- It simplifies reality so it’s easier to grasp / understand
- Has relevant aspects to analyze the issue at hand (not “complete”)
- Helps discussion, debate and to make changes
Examples of Models:
 Types of Models

Mental Model Causal Model Stock-Flow Diagram Simulatable Model

• In our mind • Causal Loop Diagrams • Stock and Flow • Stock and Flow Structures
• Maybe in text • Shows causal Structures • Has data and formulae
• pay attention only to relationship between • Without data or • Detailed mode structure
certain parts that matter elements with feedback mathematical equations
to us / familiarity

Degree of ambiguity reduces—more clear and complete

Increasing explicitness—easier to explain and debate

Eg: Consider a town with fixed geographical area. Aside from attractive
location, climate, etc., the principal determinant’s of population growth in town
are housing availability and job availability

Housing Natural Job

availability attractiveness availability

Population
A Detailed Example
Modelling Systems Using Stocks & Flows

● Stocks describe the state of a system

● Can be measured in a snap-shot of the system

World Dynamics: Mini World

En-Roads: Climate Simulator
En-Roads: Climate Simulator
World3 Model: Limits to Growth
Scenario 1: Standard Run
Scenario 2: Double the Resources
Scenario 9: Self-Imposed Limits
and Use of Technology
 Using Simulation and Modeling

● Formal Specification of a development situation

● Simulation can unfold long term dynamics
● Policy interventions can be tested using sensitivity analysis and
leverage points can be identified
● Formal simulation model forces one to explicitly state all
assumptions
● Leads to better communication of assumed system structure
 When is System Dynamics Not Useful?

● Static scenario
● Linear systems without feedback
● Here and now, not long term dynamics
○ Risks of malicious AI on election security
● Depends on timeframe and scope
○ Dividing states for better administration and development
● Detailed models capturing individual/spatial variations
○ Agent based models
 Course Expectations

● SD is not a magic wand to solve your problems

○ you have to do all the work
○ In fact, more work than if you were not using SD
○ the increased effort will increase the probability of better
outcomes
● A framework to formalize your arguments (thinking it through)
○ like all good frameworks, will instill discipline and
thoroughness
○ help identify system traps and turn them into opportunities
(turn vicious cycles into virtuous cycles)
 Logistics

● Always bring paper-pen to the class

● Always write down the answer when asked
 Text-books/References

● News sources (the World as your worksheet)

● Sterman, John D. Business Dynamics : Systems Thinking and
Modeling for a Complex World, Indian Edition : 2010
 Communication

● Moodle
● Matter of common interest on common channels, private interest
on chat
● All emails should have CS 752 or IE 604 as first part of subject
● Email header should be as specific as possible – say ‘CS 752
Midsem’
● Office Hours
● After class, 12.30-1 in the classroom
● Any time by email appointment
 Two-way Feedback

● Feedback to the students

○ HW, Quizzes, Exam
● Feedback from the students
○ During the class – ask if something is not clear, as it will help
pace the lecture
○ While reviewing a lecture – make a list of questions, note
down inaccuracies
○ Use email for subjective feedback, always ask technical
questions in person or on newsgroup
 Academic Honesty

● Absolute must
● You will be allowed to discuss things but must credit the source if
the idea/wording is not yours and not get marks for that part
● Grade penalty for cheating in homework
● Fail grade for cheating in exam/project
● We will work with you to help you achieve your potential
and put in whatever effort needed
● But our effort will be proportional to your effort
 Evaluation (TBD)

● Proctored Exams/Quizzes:
○ Endsem (written + modeling test) 35, Midsem 20, Quizzes 15
● 30% Project/class participation/ HW
● Semester Long Project: 2 components
○ Group project : Select an individual topic and build on it
● Relative Grading: But all can get AA or FF

Continuous participation expected

 Project

● Pick a topic of your choice, Preferably current such as:

○ Climate change,
○ Electric Vehicles,
○ Bottled Water,
○ NREGS expenditure,
○ Pervasive malnutrition,
○ Powai lake eutrophication,
○ Academic dishonesty in IITB
○ Liquor prohibition
○ GM Mustard
○ Cryptocurrency
○ Impact on earth of AI/ML resource consumption
○ Justice system and jail overflow
○ Communal Harmony
● Anything else is also fine
○ Can be based on primary (survey, interview etc.) or secondary (published)
data
 Project

● Every week, read at least one article (any source – can be book
chapter also)
○ note the key points: what arguments will you present to
others
○ update your SD model
○ maybe alternate between, +ve and –ve articles
● Present the model at the end of the semester and submit a
report
○ submit bi-weekly progress reports
○ will have mid-semester project viva/feedback
 Syllabus

1. Introduction: Event based vs. System Based view of the world: news paper
articles and our reaction to them
2. Causal Loop Diagrams: Causality, Causal Chains, Causal Loops, Causal Loop
Diagrams, loop polarities
3. Stock and Flow: stock, rate of change, conservation of stock, system boundary
- sources and sinks for flows,Software Support
4. Computer Simulation: Vensim, notations, balancing the units
5. Playing with Vensim: what happens to your stocks as time progresses
Advanced Issues
6. Equilibrium Diagrams
7. Fundamental modes of dynamic behavior: Exponential growth, Goal seeking,
oscillation, S-shaped growth
8. Nonlinear relationships, Delays, and, Qualitative variables: Table functions,
multipliers
9. Validating the model
10. Limitations of SD
Case Studies