Thinking tools
Second-order thinking
Keep asking "and then what?" until you have looked at the consequences of the consequences.
Howard Marks, The Most Important Thing (2011), where he calls it "second-level thinking"; revisited in his 2015 Oaktree memo "It's Not Easy".
First-order thinking stops at the immediate effect: this policy will cut the problem, this feature will raise engagement. Second-order thinking follows the chain further. How will people respond to the change, what will they do differently, and what does that response cause? Marks applies it to markets: a good company is a bad buy if everyone already thinks it is great and the price shows it.
It works because people and systems adapt. Any rule that pays for an outcome invites people to produce the measured outcome by the cheapest route, and that route is often one the designer never pictured. The first effect is usually the one you intended. The later effects are where the surprises live.
The chain never ends, so you have to stop somewhere. The skill is spotting the one or two downstream effects big enough to change the decision, then acting. Endless speculation about tenth-order effects is a way of avoiding a decision while looking thoughtful.
Example
In 1902 the French colonial government in Hanoi, worried about plague, paid a bounty of 1 cent for each rat tail handed in. On the peak day hunters brought in more than 20,000 rats. Then officials began to see tail-less rats in the streets, and health inspectors found people on the city's outskirts breeding rats for their tails. The historian Michael Vann, who found the records in the French colonial archives, describes how the bounty was cancelled.
Where it applies
Any metric you reward becomes a target, so ask how people (or an optimiser) will game it before you ship it. A model fine-tuned to maximise a grader's score will find the grader's blind spots.
Think about it
Pick a rule or incentive you introduced recently: what will people do differently because of it in six months?
Applied it to something real? Log it.
Test yourselfWhat question drives second-order thinking?
"And then what?" You ask how people and systems will respond to the first effect, and what that response causes.
Related models
- InversionWork a problem backwards by asking what would guarantee failure, then make sure you avoid it.
- One-way and two-way doorsMake reversible decisions quickly and save slow, careful deliberation for the ones you cannot undo.
- Pre-mortemBefore you start, imagine the project has already failed and ask everyone to explain why.
- Unintended consequencesPurposeful actions produce outcomes nobody planned, because the people and systems acted on respond in ways the plan never modelled.
- Goodhart's lawWhen a measure becomes a target it stops being a good measure, because people optimise the number instead of the thing it stood for.