The book "Thinking in Systems" is recommended whenever the topic of "Systems Thinking" comes up. With 200 pages and its subtitle "A Primer" it is suggested as a great introduction. I found it shallow even for 200 pages but maybe with a PhD in computer science, I'm not the right target.

Here comes my take on a quick introduction.

My Primer on Systems Thinking

A system is a set of parts and connections which fulfill a purpose together. By definition, if you take out any part or connection then the system cannot fulfill its purpose anymore.

Many system contain cycles, usually called feedback loops. feedback loops can have a balancing or reinforcing effect. A reinforcing loop makes a change result in more change. For example, the more people there are the more babies are born. A balancing loop makes a change result in a reverse change. For example, the more people there are the more will die.

One interesting aspect of a system is that we cannot apply a "divide and conquer" strategy to optimize it for its purpose. If you improve a single part then the system as a whole might suffer. Likewise, sometimes you should make one part intentionally worse to improve the system as a whole.

You can model and simulate a system in a computer. This is helpful because the behavior of a system can be unintuitive because feedback loops can be non-linear and human intuition is easily fooled by that. Another unintuitive aspect are delays in the connections. Depending on the delay, a balancing loop can either keep a system in equilibrium or oscillate out of control.

This is practically all you need to know about systems (as defined in the Systems Thinking world). You can now try to model your organization, body, family, or whatever in this framework.

Nicky Case made LOOPY which is fun to play with. For more serious use, the phrase to search for is system dynamics.


Book Contents

The book consists of three parts. The first part provides definitions like the ones above but with more examples. It shows simple systems which contain only few parts and yet result in surprisingly complex behavior.

The second part explores what high level behaviors systems have. For example, you can model the tragedy of the commons well in the Systems Thinking framework. For each such "trap", Meadows provides a short suggestion on how to resolve it. It might be fruitful to look for them in your vicinity and consider the suggestion. The answer is usually simple but not easy to implement.

The third part of the book describes Donella Meadows approach on how she tries to improve or fix systems. In her experience, the available levers to improve a system in the order of increasing effectiveness are:

  1. Numbers
  2. Buffers
  3. Stock-and-Flow Structures
  4. Delays
  5. Balancing Feedback Loops
  6. Reinforcing Feedback Loops
  7. Information Flows
  8. Rules
  9. Self-Organization
  10. Goals
  11. Paradigms
  12. Transcending Paradigms

The first six items on this list are the structure of a system, the nodes, edges, and properties. She considers these less important because changing them has either little impact or the change is practically impossible. However, sometimes there are exceptions. Sometimes structural things are easy to change and have a large impact, so one should explore the possibilities. The aspects from information flows to goals are not too hard to change and provide significant impact, so in Meadows' experience there is often leverage there. A paradigm shift is the most powerful tool she describes. There are certainly examples where a paradigm shift was caused with little effort and huge impact. However, I do wonder if these are rather coincidences instead of tools which can be applied to change a system. She does not write much about this topic and hints at Thomas Kuhn instead. Her last lever "transcending paradigms" sound to me like Buddha enlightment mumbo jumbo without any practical advice. Meadows believes Systems Thinking makes you a wiser person in general.

The last pages of the book contain more of such impractical advice. The last section title is "Don't Erode the Goal of Goodness" and the closing paragraph is:

Systems can only tell us to do that. It can't do it. We're back to the gap between understanding and implementation. Systems thinking by itself cannot bridge that gap, but it can lead us to the edge of what analysis can do and then point beyond–to what can must be done by the human spirit.


Systems Thinking provides a framework to model a system holistically. If it works well, it will give you simple answers on what to change. It provides no guidance on implementing the change. There seems to be no guidance on how to model a system well. How detailed should the model be? This book does provide no answer to that.

So when should you invest the time to model and simulate a system? The primary benefits seems to be to reveal bias. If you deal with a problem where all efforts to fix it result in counter-intuitive effects then Systems Thinking can lead to new ideas and may explain why the efforts failed. Donella Meadows has a background in environmental science which definitely exhibits this symptoms.

I think about software development a lot. There are similar symptoms here. For example, an organization introduces Scrum to become more agile but the effect is that development slows down. The idea of DevOps, where developer teams can release directly to customers without a separate Ops gatekeeper, is an instance where a delay is reduced with often good results. Requiring developers to write lots of unit tests slows development but the overall system becomes faster.

I believe Systems Thinking is worthwhile when you ran out of ideas. It can give you new ideas. However, there is no guarantee that these new ideas are good because you cannot know if your system model is good.

The book is a nice introduction to the mindset of Donella Meadows and Systems Thinking. For math-minded people, I can imagine a better primer but I do not know if it exists.

Discussion on agrees with my conclusion.


Book review: A shallow introduction to Systems Thinking.