Simulating Complex Systems with Feedback Loops

51 concepts44 papers1 proposal

When we try to understand complex socio-economic, industrial, or environmental systems, our instinct is often to build a checklist of independent factors. For instance, to boost technical innovation within an industry, we might list funding, talent, and policy as separate inputs.

However, real-world systems rarely behave linearly. An increase in funding might attract talent, which drives innovation, which in turn attracts more funding. This is circular causality. When variables are deeply intertwined, changing one component can trigger a cascade of unintended consequences across the entire network.

To navigate this complexity, researchers use a modeling paradigm that treats systems not as static collections of variables, but as dynamic networks of feedback loops, accumulations, and time delays. Rather than asking "What factors cause an outcome?", this approach asks "What is the underlying structure that generates this behavior over time?" As shown by Matsuo et al., shifting from static rules to dynamic simulations allows observers to see exactly how changes propagate through an environment, building an intuitive grasp of long-term consequences.

Mapping Structure with Causal Loops

The first step in understanding a complex system is mapping its qualitative relationships using Causal Loop Diagrams (CLDs). In a CLD, variables are connected by directed arrows that indicate the direction of influence. A positive link means the variables change in the same direction (an increase in $A$ leads to an increase in $B$); a negative link means they change in opposite directions (an increase in $A$ leads to a decrease in $B$).

These individual links combine to form two foundational types of feedback loops that govern all systemic behavior:

  1. Reinforcing Loops (Positive Feedback): These loops amplify change, driving exponential growth or accelerating decline. For example, Qi et al. mapped how distressed "zombie companies" survive through a self-reinforcing cycle where external government subsidies offset operational losses, creating an artificial stability that prevents necessary market exits or structural reforms. Similarly, Klement et al. demonstrated how data quality and knowledge sharing form a reinforcing loop that continuously drives both sustainability and operational resilience in supply chains.
  2. Balancing Loops (Negative Feedback): These loops seek equilibrium, resisting change and forcing the system toward a target state. In environmental management, Liu et al. demonstrated how human water demand and ecological constraints form balancing loops, where resource depletion eventually exerts downward pressure on economic growth to prevent total collapse.

To build these diagrams accurately, Buys et al. highlighted the value of group model-building workshops, where diverse stakeholders collaboratively map out these causal loops to uncover hidden systemic interactions. By tracing how shifts propagate through these networks, researchers can reveal the underlying structural drivers behind complex trends, such as the coordination between agricultural e-commerce and regional logistics studied by Yang.

Quantifying State and Change: Stocks and Flows

While causal loops provide excellent qualitative insights, simulation requires mathematical precision. This is achieved by translating causal loops into Stock and Flow Diagrams. This mathematical distinction is what allows the model to simulate behavior over time.

  • Stocks represent accumulations. They describe the state of the system at any given moment (e.g., inventory levels, accumulated capital, or total carbon emissions). Stocks act as the system's memory and can change only over time.
  • Flows represent the rates of change that alter those stocks (e.g., production rates, burn rates, or adoption rates).

An intuitive way to visualize this is a bathtub. The volume of water in the tub is the stock. The faucet represents the inflow, and the drain represents the outflow. The water level can only change if there is an imbalance between the inflow and outflow rates.

 [ Inflow Rate ] ---> [___ STOCK ___] ---> [ Outflow Rate ]

This distinction is vital for identifying operational vulnerabilities. For example, Gu et al. and Gu et al. utilized stock-flow structures to simulate multi-tier reverse supply chains. By modeling how consumer collection rates (flows) affect disassembly inventories (stocks), the authors quantified the financial damage of supply disruptions, revealing why early-stage disruptions inflict disproportionate damage across the entire network.

In urban planning, Li et al. coupled stock-flow models with spatial expansion data to simulate how construction land structures evolve in shrinking cities, proving that growth-focused planning can be counterproductive when underlying population stocks are declining.

The Friction of Reality: Time Delays

In a simple system, cause and effect happen almost simultaneously. In complex systems, however, there is often a significant gap between an action and its visible result—a time delay. Delays introduce friction and instability into a system, frequently causing decision-makers to overcorrect or abandon effective strategies prematurely.

Consider policy interventions. Chen et al. applied this modeling approach to cross-border e-commerce and identified a substantial policy lag between the introduction of regulatory support and actual infrastructure development. Without accounting for this delay, policymakers might assume an intervention has failed and implement a new, conflicting policy.

Organizational workflows suffer from similar frictions. Liu et al. modeled open-source intelligence processes, showing how information delays and accumulation across functional departments bottleneck the speed and quality of intelligence production.

Why It Works: Uncovering Leverage Points

The ultimate power of feedback-loop simulation lies in its ability to run "what-if" scenarios over multi-year or multi-decade horizons. This allows researchers to find leverage points—places in a system where a small structural shift can yield large, systemic improvements.

Often, the most obvious solution is not the most effective one. For instance, in regional financial cooperation, Zhu et al. simulated various policy levers and discovered that optimizing legal and business climates was far more effective at driving long-term cooperation than simply injecting capital through loans.

Furthermore, these models allow us to navigate complex trade-offs where multiple distinct systems are coupled together:

  • Energy and Carbon Markets: Zhang et al. and Zhang et al. simulated the non-linear feedback between regulatory carbon quotas, renewable portfolio standards, and electricity prices, mapping exactly how short-term price increases eventually stabilize as infrastructure investments catch up. On a broader scale, Chen et al. used dynamic integrated assessment modeling to capture the multi-decade feedback loops between carbon capture infrastructure and regional GDP growth.
  • Socio-Technical Systems: Bian et al. argued for treating healthcare environments as interconnected networks of technology, policy, and human behavior, using systems thinking to address complex issues like workforce burnout.
  • Industrial Innovation: Yu et al. and Yu et al. mapped the innovation ecosystems of manufacturing and education, demonstrating how internal management processes are linked to external environmental pressures through causal feedback loops.
  • Security and Technology Integration: The authors of Zheng et al. integrated these feedback models with real-time Internet of Things (IoT) and blockchain data, showing how safety investments and operational efficiency interact over time to create robust early-warning security systems.

By formalizing the hidden feedback loops, accumulations, and delays that govern our world, this approach transforms complex systems from unpredictable black boxes into understandable, manageable structures.

Key Foundational Work

Towards a multi-dimensional framework for characterising and comparing porkchains (2008) — contributed 1 concept elements to this direction.
A Learning Support Method in Qualitative Simulation-Based Economic Education (2005) — contributed 1 concept elements to this direction.
History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_system_dynamics_modeling_64ca6971,
  title = {system dynamics modeling},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/system-dynamics-modeling/rdr_64ca6971412b40cea11475dee287cb9b},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-20},
}

This exact version (version 1, frozen text):

@misc{lacuna_system_dynamics_modeling_64ca6971_v1,
  title = {system dynamics modeling},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/system-dynamics-modeling/rdr_64ca6971412b40cea11475dee287cb9b/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-20},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake