Meadows on why systems surprise us
📚SOURCE
💎PERMANENT NOTES
🧵SYNTHESIS
Paraphrase, not quotation: chapters 4 and 5 of 📕 Thinking in Systems.
Chapter 4 is a list of reasons our mental models fail: we watch events and not the behaviour over time, we think in straight lines, we draw boundaries where it is convenient, and we forget delays. The one that stayed is bounded rationality. Each person in a system acts sensibly on the little they can see, and the sum of sensible acts can be a result nobody wanted.
Chapter 5 turns that into a catalogue of traps. Policy resistance: several actors each pull a stock toward their own goal, so every push is cancelled. Drift to low performance: the standard is set by recent results, so each bad month quietly lowers the bar.
Her claim, and now mine: put different people in the same seats and you get the same behaviour. Change the information or the goal and the same people behave differently.
✨KEY IDEAS
People act reasonably on the partial information in front of them; the trouble is what the information leaves out.
A system boundary is a choice made for the question in hand, not a fact about the world.
In policy resistance each side works harder and the stock stays where it was.
When the standard follows performance down, the fix is to anchor it to the best results and not the latest ones.
The way out of a trap is usually to change the goal or the information, not to push harder inside it.
📝READING NOTES
Why I read this half against my own team
The first note was about bathtubs. This one is about why clever people standing around a bathtub still get it wrong, and I read it with the launch plan open in the other window.
Chapter 4: six ways to be surprised
- Events. We explain the week by what happened in it. She wants the graph over months, and then the structure that draws the graph.
- Straight lines. Twice the load is not twice the trouble.
- Boundaries. Every diagram ends somewhere, and the thing that bites you is just past the edge.
- Limits. There is always a limiting factor, and removing one promotes the next.
- Delays. Covered in 🧩 Delays in feedback.
- Bounded rationality. Nobody sees the whole. Everybody optimises their corner.
Chapter 5: the traps I can name people in (and should not)
Policy resistance. Joachim wants the date. Lucía wants a stable sprint. Nadia wants a test that means something. I want all three and say so in three different meetings. Each of us is rational. The launch date has not moved in six weeks and neither has the risk.
Drift to low performance. Tomorrow I am due to agree load test volumes with Nadia. She thinks the plan is too gentle. The volumes on the table are based on what passed last time, which is this trap exactly: the bar is last quarter's result.
Blame the structure can become a way of never saying a hard thing to a person. Meadows does not claim people never fail. She claims structure explains more. I am keeping the word more in the title of the permanent note on purpose, and the counter-argument lives in 🧠 Most problems are structure, not people.
What it produced
💎 The structure of a system explains more than the people in it, written the same morning, before brunch.
Added 19 March: the retro today was blameless and it was not kindness. With the names taken out, the timeline showed two information gaps that any team would have fallen into.
Questions it left me with
- How would I know it was a person and not the structure? I do not have a test, only a preference.
- If the load test standard has drifted, who owns the anchor? Nadia, me, or nobody, which is the usual answer.
- She says redesign the information. What is the smallest piece of information my team is missing that Kestrel's dispatchers already have?
Reading record
- Section read: Chapters 4 and 5
🗣️IN SHORT
- In : Thinking in Systems, chapters 4 and 5
- Meadows argues that : systems surprise us because each actor decides sensibly on partial information, and a small number of recurring structures turn those decisions into bad outcomes
- by stating that : policy resistance, drifting standards and the other traps reappear with any cast of people, and loosen only when goals or information flows are redesigned