System Dynamics: Systems Thinking and Modeling for a Complex World

Main ideas, concepts, and lessons

1) What the session covers (and key framing)

• The talk is a 3-hour overview of systems thinking and system dynamics for complex problems.
• A key emphasis is the subtle difference between:
  • System dynamics: a set/methodology for modeling complex systems.
  • Systems thinking: the broader conceptual field/way of understanding systems.
• Because the session is short, it doesn’t deeply dive into all tools and technical methods, but it lays out foundational concepts.

2) Who the speaker is (James Payne) and why he’s interested

• James Payne is a second-year PhD student in MIT’s System Dynamics group.
• Background (industry/engineering/management):
  • ~10 years as a chemical engineer (GE Hitachi Nuclear Energy)
  • Industrial engineer / Lean Six Sigma (Inmar; reverse logistics)
  • Product manager (Hanesbrands / Maidenform / Playtex brands)
• Why system dynamics drew him in:
  • In interviews, he could explain a very diverse career (nuclear reactors → bras) in a way that made sense when people connected it to system-level thinking.
• Research interest:
  • Behavioral operations management (Behavioral OM):
    • Focuses on how human decision-making within supply chains creates system behavior.
    • Core principle: “the presence of people is not a bug, it’s a feature.”
    • The goal is to plan around and with people, not pretend they’re not there.

3) MIT System Dynamics group: people and field background

• Faculty mentioned:
  • John Sterman, Nelson Repenning, Hashim Ramdad, David Keith
• Core textbook recommendation:
  • John Sterman’s Business Dynamics (especially first two chapters for fundamentals).
• Brief history of system dynamics:
  • Origins at MIT; rooted in control theory.
  • Central figure: Jay Forrester
    • Developed Industrial Dynamics (1958), described as seminal.
    • Earlier computing work (including reference to Whirlwind, an MIT digital computer).
  • Example development story:
    • GE appliance plant saw a 3-year boom/bust employment cycle.
    • Forrester built early system dynamics models to explain why.
  • Larger-scale work:
    • System dynamics applied to social systems via the World2 simulation.
    • Led to The Limits to Growth—not intended as precise future prediction, but as commentary on social structure and resource renewal/limits (and noted as controversial due to misinterpretation).

4) Systems thinking vs linear “open-loop” approaches

• Common linear workflow shown:
  • Identify problem → gather data → evaluate alternatives → select solution → implement
• Why this is incomplete (the systems view):
  • Changing system state changes:
    • Implementation/control processes
    • Goals
    • Other decision-makers (suppliers, customers, other external actors)
  • Results include:
    • Side effects (reframed as “effects you hadn’t thought about yet”)
  • Delays are crucial:
    • Changes today can loop back later, affecting the original issue again.
• Core maxims:
  • No decision exists in isolation.
  • No process exists in isolation (only boundary choices differ).
  • Feedback closes the loop; system dynamics helps “close” it.

5) Definition of “system” and properties of social-economic systems

• A system: a set of interdependent parts with a common purpose.
• Boundaries must be defined (otherwise everything becomes the system).
• Social-economic systems are described as:
  • Highly complex, dynamic, tightly coupled
  • Governed by feedback
  • Non-linear
  • Limited information and ambiguity
  • Often more complex than physical systems

6) Methodology: mental models + simulation (and what models are for)

• Central process:
  • Elicit mental models:
    • Understand why people make decisions they consider rational.
  • Place these mental models inside a larger system context and observe feedback effects.
• Simulation:
  • Used because it’s cheaper/faster than real-world experimentation.
  • Also teaches by showing people how their mental model “fits” and how it interacts with feedback.
• Important modeling lesson:
  • Simulation’s purpose is not to be perfectly “right.”
  • A model should be useful:
    • Focus on high-leverage policy choices
    • Identify structures that drive modes of behavior, not necessarily match every real-world detail exactly.

7) Three core points of systems thinking / system dynamics

1. Structure generates behavior
   • People’s actions are feasible and rational given the structure they operate in.
   • “Structure” includes:
     • Physical structure
     • Information availability
     • The “structure in people’s heads” (decision logic / mental models)
   • Mentioned control-theory framing (e.g., “tiny PID controllers” idea).
2. Fundamental attribution error
   • People initially blame individuals rather than system structure.
   • Example: road-rage reaction after being cut off → assumes the other driver is “bad,” ignoring underlying rationality/constraints.
   • Emphasized in their group culture:
     • Assumption that others are intelligent, capable, caring, acting with integrity, and want to learn.
     • Therefore, focus on changing the system, not just the person.
3. Events → patterns → structure
   • Events are visible occurrences.
   • Repeating events form patterns of behavior.
   • Patterns arise from deeper structure (including physical/information/mental-model structure).

8) How mental models are hard to change

• Even if you tell people to incorporate feedback, it’s incredibly difficult to learn and change behavior in dynamic complexity.
• Reasons referenced:
  • Limited information
  • Time delays
  • People can only decide based on what they can observe/act on.

9) Usefulness limits: forecasting precision vs projecting behaviors

• System dynamics is framed as:
  • Poor for precise point predictions far into the future
  • Good for projecting modes of behavior
    • Example idea: predicting oscillatory behavior characteristics (amplitude, periodicity) rather than exact future values at a time point.
• The key value:
  • Identify structures with policy impact on behavior over time.

---

Detailed bullet list: Tooling and modeling concepts taught

A) Spiral/iterative modeling approach

• System modeling is iterative; no single “right” model at the start.
• The process is like a spiral:
  • Choose a reference mode (a behavior pattern you observe)
  • Build a model that reflects it
  • Add more reference modes / new structure when needed
  • If adding structure breaks alignment, rework earlier assumptions
• Simulation helps manage multiple candidate behaviors stories:
  • If you get multiple stories/outcomes from interviews:
    • Model one mode first
    • Then model another
    • Then check how they can be unified/related (overlap/transition)

B) Causal links and causal-loop diagrams (CLDs)

• Causal link definition:
  • “If one thing changes, another thing changes.”
• Examples of clearly signed causal links:
  • Production ↑ → inventory ↑
  • Shipments ↑ → inventory ↓
  • More sales force → orders booked ↑
  • Price ↑ → orders ↓
  • Births ↑ → population ↑
  • Deaths ↑ → population ↓
• Handling ambiguity:
  • If a relationship could be either positive or negative, it implies missing structure in the middle.
  • The loop isn’t “single-step”; you need intermediate mechanisms.

C) Distinguish causality vs correlation (as operationalized)

• The speaker stresses that diagrams should reflect causal reasoning, not mere correlation.
• Example used:
  • Ice cream sales and murder rates rise together (correlation).
  • The “real” drivers are explained by an intermediate variable (temperature/people interactions).
• In social systems, causality is acknowledged as more “operational/assumption-dependent,” but still requires consistent directional coupling for usefulness.

D) Causal loops: reinforcing vs balancing loops

• Build loops by tracing variables around until you return to the starting variable.
• Reinforcing vs balancing rule of thumb:
  • Count negative links in the loop:
    • Even number of negatives → reinforcing loop
    • Odd number of negatives → balancing loop
• Reinforcing loop example (described):
  • Coaching time enables skills growth → improves customer satisfaction → reduces complaints → frees manager time → more coaching time.
• Balancing loop example (described):
  • Market attractiveness ↑ → more competitors ↑ → competitive effects push price down → profits down (balancing tendency).

E) Stocks and flows (compartmental models)

• Stock:
  • A quantity with memory (accumulation over time).
• Flow:
  • Rates that increase/decrease the stock.
• Modeling metaphor:
  • “Bathtub” / bucket with inflow and outflow:
    • You can’t change the stock instantly without affecting inflows/outflows.
• Examples mentioned:
  • Stock-like: number of employees, inventory units, greenhouse gas accumulation.
  • Other categorization examples:
    • Interest rate treated as a stock in the talk’s framing because it’s a sticky price (not a pure flow rate, despite the word “rate”).
• The speaker contrasts:
  • Not all “system dynamics” work is limited to stock-and-flow diagrams, but they are common and foundational.

---

Simulation / experiential component

• A major part of the afternoon is a management flight simulator using system dynamics—described as their “hands-on” activity.
• The simulator is followed by discussion tying simulation lessons back to earlier concepts.
• Participants’ simulation theme:
  • Fishbanks: deep-sea/coastal fishing enterprise management.
  • Goal: catch fish and make money using learned understanding of systems behavior.
• A Q&A point during the session:
  • A team found “running too fast” caused problems—suggesting they didn’t understand the model equations/feedback timing.
  • Information is imperfect:
    • They discuss running again with more known information and reference the Beer Game example (information matters, but structure still drives outcomes).

---

Additional examples used to reinforce lessons

• Oil prices example:
  • Illustrated with event-by-event narratives (spikes/drops) vs deeper boom-bust pattern arising from structure.
• Vasa ship example (unintended consequences):
  • Late design changes increased sails/cannons/lodge size → hydrodynamic instability → ship capsized/sank.
  • Used to emphasize: there are no “side effects,” only unplanned effects.

---

Resources listed (courses, articles, books, websites, software)

MIT courses mentioned

• 15.871 Introduction to System Dynamics
  • Emphasis: get systems thinking in your head; causal loop diagrams and reasoning.
• 15.872 (continuation / modeling tools)
  • More realistic scenarios; modeling with software; “training wheels off.”
• 15.873 (newer, between 15.872 and 15.872/positioned)
  • Emphasis toward business and policy.

Books mentioned

• Business Dynamics (John Sterman) — foundational textbook.
• The Fifth Discipline Fieldbook — summary of systems thinking (particularly part two).
• The Limits to Growth — history of overshoot/collapse/resource limits; clarifies “mode of behavior” vs exact prediction.

Articles (general references)

• System Dynamics 660: The Path Forward
• An article emphasizing: system dynamics is not compartmental models; it is a mode of thinking.
• Making the Numbers
  • Connects an operations management concept (“capability trap”) to valuation of stocks.

Other websites / self-study resources

• Creating Creative Learning Exchange (K-12 focused; animations/tools; mobile apps)
• Tom Fiddaman’s MetaSD (example-driven interactive modeling)
• Self-study system dynamics site (dense but structured; free/open)
• Journal club and MIT seminar/discussion structure:
  • Friday/seminar cadence on MIT’s fourth floor (mentioned as a plug).

Software tools mentioned

• Vensim (free academic license; MIT email may unlock more)
• Stella Architect
• NetLogo (agent-based modeling option; not central in their group but used by speaker)
• General note: system dynamics can be done with many tools; even Excel or a whiteboard can get you started.

Policy simulator example

• En-ROADS:
  • Climate policy simulator with underlying compartmental climate modeling.
  • Used to discuss policy choices and eliciting shared mental models.
  • Emphasized: again not for precise prediction, but for policy exploration.

---

Speakers / sources featured (as named in the subtitles)

Speaker

• James Payne (MIT System Dynamics group; main presenter)

Sources / referenced people and institutions

• John Sterman (faculty; author of Business Dynamics)
• Nelson Repenning (MIT faculty; named)
• Hashim Ramdad (MIT faculty; named)
• David Keith (MIT faculty; named)
• Jay Forrester (originator of industrial dynamics; central history figure)
• Club of Rome (referenced via Forrester’s World2 work)
• MIT System Dynamics group (institutional source)

Books / documents referenced

• Business Dynamics (John Sterman)
• The Fifth Discipline Fieldbook (systems thinking)
• The Limits to Growth (Limits to Growth; World2 context)

Software / tools referenced

• Vensim
• Stella Architect
• NetLogo
• En-ROADS
• (Simulation/game) Fishbanks
• (Reference example) Beer Game
• MetaSD (Tom Fiddaman’s site)

Other websites / organizations referenced

• Creating Creative Learning Exchange
• System Dynamics Society
• Dewey Library (where books are suggested to be found/borrowed)