Thinking in Systems

Read 16 to 22 Feb 2026. Donella Meadows, Thinking in Systems.

The central argument is that almost everything we encounter is a system. Economies, organisations, conversations, immune responses. Once you see it that way, you stop asking what caused something and start asking what loop is this part of.

Stocks are what accumulates. Flows are what change them. Feedback loops are what connect them. That’s the whole grammar.

Feedback loops and delays

Reinforcing loops amplify things. Balancing loops regulate things. Every single one has a delay. The delay is where all the trouble happens. You act, the system responds, but the response takes time. So you overshoot. You overcorrect. Most policy failures aren’t about bad intentions. They’re about acting before the feedback arrives.

Systems nest

A team is a system. That team is inside a company which is inside an industry which is inside an economy. There’s no clean boundary anywhere. You draw a boundary because you need to think about something specific, not because the boundary is real. The interesting behaviour often lives at the boundary you chose not to draw.

Models, not truth

We never see a system directly. We see our mental model of it, which is always incomplete and shaped by what we already believe. Meadows keeps reminding you of this: hold your mental model lightly. It’s a hypothesis, not a fact.

The part that stuck with me is that this applies to “truth” itself. What most people call truth is a socially constructed model that has been fed back into itself long enough that it feels solid. It’s still a model.

Language shapes what you can see

If you have no word for a concept, you’ll struggle to perceive it or reason about it. Different linguistic communities literally structure the same problem differently. The categories available to you constrain the systems you can see.

This feels important for technical work. The vocabulary a team uses shapes what problems they can even name.

What the book asks you to do

Not to find better levers to pull. To learn to see structures rather than events. Events are the surface. The loop underneath is what produces the events, and the loop is what needs to change.