Systems Dynamics

MIT electrical engineer. Pioneer of digital computer development — worked on the Whirlwind I computer (the first real-time computer, built for the US Navy). From digital computers to modeling social systems: the transition happened when he realized that the same feedback dynamics governing machines governed organizations, economies, and ecosystems.

The insight: Industrial firms behaved like dynamic systems with feedback loops — production decisions influenced market conditions, which influenced future production. Traditional business management couldn't see this because the delays were too long and the feedback too complex for unaided intuition. Computer simulation could.

The founding works:

Industrial Dynamics (1961) — first application to industrial firms and business cycles

Urban Dynamics (1969) — applied to urban planning and decay

World Dynamics (1971) — the model that became The Limits to Growth

The fundamental building blocks. Everything in SD is a stock (accumulation) and flows (rates of change into and out of that stock).

Stock: a quantity that accumulates over time. Population, capital, inventory, debt, water in a reservoir. Stocks change only through flows.

Flow: the rate at which a stock changes. Births/deaths, investment/depreciation, purchases/consumption. Flows are rates per unit time.

The key rule: stocks change only through flows. If you can identify the stocks and flows in a system, you have a structural model of it.

Stocks influence their own flows through feedback loops. Two types:

Reinforcing (positive) feedback loop:

Stock increases → flow increases → stock increases faster → flow increases even more. The loop amplifies change. Example: compound interest, population growth, viral social media spread.

Balancing (negative) feedback loop:

Stock increases → flow decreases (or opposite direction) → stock decreases toward goal. The loop stabilizes. Example: thermostat, supply-demand pricing, predator-prey cycles.

Real systems have multiple interlocking reinforcing and balancing loops. The behavior of the system emerges from which loop dominates at which time.

The most counterintuitive element. Delays between a cause and its effect cause systems to overshoot and oscillate. The classic example: a hot shower where you turn up the cold tap, wait 10 seconds, feel no change, turn up more, then get scalded.

In social systems, delays are everywhere:

Investment takes years to produce capacity

Population responds to conditions with a generation lag

Policy changes take time to filter through institutions

Perceptions of environmental change lag behind actual change

Forrester's key claim: delays cause instability. Systems with long delays and strong reinforcing loops inevitably overshoot and collapse — not because of bad decisions, but because of system structure.

Visual representation of feedback structure. Shows stocks, flows, and the causal links between them. The qualitative first step before building a quantitative model.

From MIT's System Dynamics in Education Project materials:

S-Shaped Growth:

Reinforcing loop dominates early → exponential growth

As stock grows, a balancing loop activates (resource scarcity, competition)

Growth slows, stabilizes at carrying capacity

Classic pattern: population approaching the limits of its environment

Overshoot and Collapse:

S-shaped growth continues past the sustainable level

Stock overshoots carrying capacity

Resources depleted beyond recovery

Balancing loop eventually dominates → collapse

Examples: Goldfield, Arizona (gold rush → boom → bust), deer population destroying its own food supply, fisheries collapsing

Oscillations:

Delays in feedback cause the system to overshoot in both directions

Stock oscillates around its target

Example: business cycles, predator-prey populations, housing markets

Forrester built World Dynamics (1971) — a global model incorporating population, capital, pollution, resource depletion, and food production. The model showed that exponential growth in population and industrial capital would eventually hit resource and environmental limits, producing collapse.

This model was the direct inspiration for The Limits to Growth (1972) by Donella Meadows and the Club of Rome — the most famous application of systems dynamics to a global problem. Limits to Growth used a more developed version of Forrester's World model and made the SD methodology accessible to a general audience.

The Limits to Growth controversy: Published in 1972, it predicted resource scarcity-driven economic decline by the mid-21st century if growth continued unchanged. Critics (especially economists) dismissed it as too pessimistic and based on oversimplified models. 50-year follow-up studies (2014, 2021) found that actual resource trends were tracking the model's pessimistic scenarios fairly closely — a significant vindication.

1. Identify the problem — what behavior are you trying to understand or change?

2. Identify the stocks — what accumulates over time?

3. Identify the flows — what causes stocks to increase or decrease?

4. Map the feedback loops — which loops are reinforcing, which are balancing?

5. Identify delays — where are the time lags between cause and effect?

6. Estimate parameters — get approximate values for rates and constants

7. Build the simulation — translate the causal structure into equations

8. Test against history — does the model reproduce observed behavior?

9. Run scenarios — what happens if you change assumptions?

10. Identify leverage points — where can small interventions produce large changes?

Developed by isee systems in 1985, making SD accessible beyond programmers. Drag-and-drop interface for building stock-flow models. Became the standard tool for SD education and practice.

Forrester's most important insight: social systems have a structure that makes them produce counterintuitive behavior. People in the system act rationally given the information available, but the system as a whole behaves in ways no individual intended.

Key counterintuitive patterns:

Action causes the opposite reaction — pushing on a system produces delayed, often opposite effects

Interventions that work short-term make long-term worse — e.g., increasing production to meet demand depletes resources faster, requiring more production

The obvious solution is often wrong — adding capacity to fix a shortage overshoots into surplus, because demand also grows in response

The place you least expect is where the leverage is — leverage points are rarely where intuition suggests

This is directly relevant to Peter Turchin's cliodynamics: the disintegrative phase isn't caused by bad actors — it's caused by the structural dynamics of the wealth pump. The system produces the crisis.

Business: supply chain management, corporate strategy, product lifecycle modeling

Urban planning: urban decay and renewal dynamics

Public health: epidemic modeling, healthcare system dynamics

Environment: resource depletion, climate modeling, fisheries management

Policy: drug policy, energy policy, economic stabilization

Psychology: cognitive-behavioral dynamics, habit formation

Formal and testable — quantitative models that can be calibrated against data

Reveals structural causes — why systems behave as they do, not just what happens

Scenario testing — can explore "what if" questions about policy interventions

Connects to intuition — stock-flow diagrams make system structure visible

Identifies leverage points — shows where interventions have maximum effect

Model complexity vs. real complexity — SD models are simplified; real systems have more variables

Parameter estimation — getting the right numbers for rates is often difficult

Structural assumptions matter most — the qualitative feedback structure is more important than precise parameters, but getting the structure wrong produces wrong conclusions

Not predictive in the narrow sense — models show how systems behave under assumptions, not exact predictions

Can be gamed — stakeholders can manipulate models to support predetermined conclusions

Yaneer Bar-Yam's complex systems science uses SD-like feedback structures but adds multiscale analysis, network theory, and phase transition analysis that SD doesn't formalize

Peter Turchin's cliodynamics is essentially SD applied to historical macrosociology — the three-compartment model (population, elites, state) with nonlinear feedback loops is a SD framework

Forrester's counterintuitive behavior — "people act rationally but the system behaves badly" — is the theoretical basis for why SDT predicts instability: not from bad actors, but from system structure

World Dynamics → Limits to Growth → Peter Turchin — the lineage of global SD modeling for civilizational analysis

Overshoot and collapse is the generic SD structure behind Peter Turchin's disintegrative phase

Forrester, Industrial Dynamics (1961)

Forrester, Urban Dynamics (1969)

Forrester, World Dynamics (1971)

Meadows, The Limits to Growth (1972) — most famous SD application

Meadows, Thinking in Systems (2008) — accessible introduction by Meadows, the student who became the master

Sterman, Business Dynamics: Systems Thinking and Modeling for a Complex World (2000) — the definitive modern SD textbook

MIT System Dynamics Group: systemdynamics.org

Psychohistory

Peter Turchin

Yaneer Bar-Yam

• Psychohistory
• Peter Turchin
• Yaneer Bar-Yam

Categories: Home › Systems

Home · Generated from MemPalace