Dave's reading notes

systems-thinking

Properties1
tagscognition, complexity, concept-epistemology, meadows, systems

Systems Thinking

Summary: A mode of analysis that focuses on how parts of a system interrelate and how the system behaves over time, rather than on linear chains of cause and effect. The central source is Meadows’s Thinking in Systems.

Sources: Clippings/Meadows-Thinking-in-Systems.md

Source pages: Thinking in Systems

Last updated: 2026-05-04

What systems thinking is

Conventional analysis isolates variables and looks for direct causes. Systems thinking asks: what is the structure that produces this behavior? Structure — the combination of elements, interconnections, and purpose — generates outcomes that are often counterintuitive and resistant to piecemeal intervention.

The key shift is from event to pattern to structure: not “what happened?” but “what is producing this kind of thing repeatedly?”

The three components of any system

  1. Elements — the visible, concrete parts (people, trees, machines, institutions). Elements are the easiest to notice and the hardest to change.
  2. Interconnections — the relationships that hold elements together (flows of information, material, energy, money). Often invisible, often the real leverage point.
  3. Function or purpose — what the system actually does (not what we say it does). Purpose is revealed by behavior, not by stated intention.

Changing purpose changes the system more than changing any element. Changing interconnections changes it most subtly but often most durably.

Feedback loops

Systems are not linear pipelines; they loop back on themselves. Two basic types:

  • Reinforcing (positive) loops: amplify change in one direction. Growth, collapse, vicious cycles, virtuous cycles are all reinforcing loops.
  • Balancing (negative) loops: resist change, seek a goal or equilibrium. Thermostats, predator-prey dynamics, budget constraints.

Most interesting behavior comes from the interaction of multiple loops — which is dominant changes over time and with scale.

Delays

When there is a gap between an action and its feedback, systems overshoot and oscillate. Actors in a system with long delays cannot respond to current conditions; they respond to old information. This is why:

  • Commodity markets boom and bust cyclically
  • Central planning fails at large scales
  • Drug policy produces the wrong interventions years after the problem has shifted

Meadows: “A system just can’t respond to short-term changes when it has long term delays.”

Goals and measurement

A system does what it is measured to do. If welfare is not measured, it is not optimized — even if it is stated as a goal. This makes measurement a political act: choosing what to track is choosing what the system will pursue. See the GNP example in meadows-thinking-in-systems.

Bounded rationality

Each actor in a system makes decisions based on local information — they cannot see the whole system. Rational individual behavior can produce irrational collective outcomes. This is not stupidity; it is a structural property of complex systems. The solution is usually better information flows, not better people.

Meadows’s intervention: “Adding or restoring information can be a powerful intervention, usually much easier and cheaper than rebuilding physical infrastructure.”

The limits of models

All knowledge is a model. Models are always simpler than reality; they are always incomplete. This does not make them useless — it makes humility essential. The error is not using models; it is forgetting that you are. See narrative-bias.

Connections to other concepts

  • chaos-and-prediction: Both Meadows and Banks (The Hydrogen Sonata) arrive at similar conclusions from different directions — complex systems are fundamentally resistant to precise prediction. More analysis does not dissolve uncertainty when initial conditions are sensitive and feedback is delayed.
  • political-manipulation: Information flows are leverage points. Suppressing or distorting information is therefore a form of systemic sabotage, not just propaganda.
  • narrative-bias: The instinct to find linear causes for outcomes is itself a cognitive model that systems thinking challenges. Events look inevitable in retrospect because we trace the single path taken, not the branching structure that produced it.
  • outside-context-problem: An OCP is the limiting case of bounded rationality — a system encountering something so outside its model that feedback loops cannot process it at all.

Graph View

Backlinks