Activate this model when facing problems embedded in complex systems (e.g., organizations, markets, ecosystems, relationships, or any situation with multiple interacting components and feedback loops).
Linear thinking systematically produces wrong answers about systemic problems. You cannot understand a complex system by analyzing its components in isolation — the system’s behavior emerges from the relationships between components, not from the components themselves.
Stock: An accumulation, what exists at a point in time. Example: Population, money in an account, trust in a relationship
Flow: A rate of change, what increases or decreases the stock. Example: Birth rate, spending rate, honesty over time
Key insight: Stocks have inertia. They change slowly even when flows change rapidly. Most interventions affect flows; most problems involve stocks. Plan for the delay between changing a flow and seeing the stock move.
Reinforcing (positive): Amplifies change, growth or collapse. Example: Compound interest; viral spread; vicious cycles
Balancing (negative): Resists change, stabilizes toward a goal. Example: Thermostat; homeostasis; budget constraints
All systems contain both. The system’s behavior at any moment depends on which loop is currently dominant.
Donella Meadows’ hierarchy, from lowest to highest leverage:
Low: Numbers and parameters. Example: Tax rates, speed limits
Medium: Sizes of stocks and flows. Example: Infrastructure investment
Higher: Structure of information flows. Example: Who has access to what data
Higher: Rules of the system. Example: Incentive structures, regulations
Highest: Goals and paradigms. Example: The purpose the system serves; the shared assumptions about how the world works
Most interventions target low-leverage points. Systemic change requires targeting high-leverage points.
Complex systems produce behaviors that no individual component has, behaviors that cannot be predicted from analyzing components in isolation.
Ask: What behaviors might this system produce that none of its parts would exhibit alone? Traffic jams emerge from individual driving decisions. Market crashes emerge from individually rational trades. Cultural norms emerge from individual choices.
Every intervention in a system produces second and third-order effects. Map at least two rounds:
What is the intended first-order effect?
What second-order effects does that produce?
What third-order effects do those produce?
Who benefits in ways they didn’t ask for? Who is harmed in ways you didn’t intend?
In systems with time delays, cause and effect are separated in time. This consistently leads to overshooting, continuing to push on a variable after the system has already begun to respond, because the response hasn’t arrived yet.
Rule: In high-delay systems, act smaller than feels necessary. Wait for the response before acting again.