Meadows on dancing with systems
📚SOURCE
💎PERMANENT NOTES
🧵SYNTHESIS
Paraphrase, not quotation: the whole essay, 📰 Dancing with Systems.
This is the essay she wrote after the leverage list, and it reads like a correction. Her generation of systems modellers assumed that understanding a system would let them predict and control it. It did not. Complex systems are nonlinear, self-organising and full of delay, so they can be understood only in general and steered only partly. What is left is to dance: to pay close attention and respond.
The advice comes as a list of practices. The first is to get the beat, which means watching how a system actually behaves over time before touching it, with data and not with opinions about it. Others: say your model out loud so that it can be shot at, protect the flow of honest information, value what cannot be counted, stay a learner.
It is a humbler text than I wanted the week before a launch.
✨KEY IDEAS
Before you disturb a system, watch how it behaves. Start from its history and its data, not from anyone's theory of it.
Put your mental model where others can see it and attack it.
Distorted, delayed or withheld information is the most common cause of a system going wrong.
Attend to what matters and not only to what can be counted.
Admitting error quickly is the only way to learn at the speed the system changes.
📝READING NOTES
Why it follows the leverage essay
📖 Meadows on leverage points hands you a ladder. This one tells you the ladder is leaning on something that moves. Read in the other order, I would have taken the ladder too literally.
The practices, in my order of difficulty
Easy to agree with, easy to do
- Get the beat.
- Use plain language. She thinks muddy words carry muddy models.
- Widen the time horizon and the circle of people you count.
Easy to agree with, hard to do
- Expose your model to criticism.
- Honour information: do not bend it, delay it or sit on it.
- Locate responsibility inside the system, so that whoever decides also feels the result.
Hard to agree with on a deadline
- Pay attention to quality when you have a number.
- Stay humble. Say I was wrong early.
- Go for the good of the whole and not for one part's score.
Reading it this particular Sunday
I fly to Rotterdam on Tuesday to sit on a dispatch floor for a morning. That is get the beat almost by accident: I booked it as a customer visit and it turns out to be the first time I will watch the thing I am launching being used at 07:00 by someone who did not ask for it.
You cannot control a system is true and a little too comfortable. Someone still has to decide the date. Meadows is writing to modellers who over-reached; I am a manager who is paid to commit. The honest version for me is narrower: I control my decision and what I show people, and I only get to listen to the rest.
What it produced
💎 I can listen to a system but I cannot control it, the same day.
Added 23 February: the launch slipped on the 12th. I wrote 🕯️ On wanting to control a launch this morning with this note open. Get the beat is what the load test was, and I had argued its volume down.
More on her in 🏛️ Donella Meadows.
Questions it left me with
- What would get the beat look like as a standing practice and not a trip? An hour a month on a customer's floor?
- Which of my numbers is standing in for something I cannot count: is open exceptions a proxy for does the dispatcher trust the list?
- Is my discomfort with this essay a sign that she is wrong, or that I like control more than I admit?
Reading record
- Read on: 2 February 2025
🗣️IN SHORT
- In : Dancing with Systems, the full essay
- Meadows argues that : complex systems cannot be predicted or controlled, only understood in outline and responded to with attention
- by stating that : nonlinearity, delay and self-organisation defeat forecasting, so the workable stance is a set of habits: observe first, expose your model, guard honest information, and keep learning