Kahneman on two systems
📚SOURCE
💎PERMANENT NOTES
🧵SYNTHESIS
Paraphrase, not quotation: parts 1 and 2 of 📕 Thinking, Fast and Slow, up to chapter 18.
Kahneman describes the mind as two characters. System 1 is fast, automatic and always on: it reads faces, finishes sentences and produces an answer before being asked. System 2 is slow, deliberate and costly, and is, in his word, lazy. It mostly endorses whatever System 1 hands it. The characters are a fiction and he says so; they are a way of talking about which answers we check.
System 1 builds the most coherent story from whatever is in front of it and does not ask what is missing. That is why a vivid description beats a base rate, and why a small sample feels as convincing as a large one.
Chapter 17 is where regression to the mean first made sense to me. Any result that is partly luck will tend to be followed by a more ordinary one, and we then invent a cause for the change.
✨KEY IDEAS
System 1 answers an easier question than the one asked, and System 2 rarely notices the swap.
Confidence comes from how coherent the story is, not from how much evidence stands behind it.
Given a description of a person, we judge by resemblance and ignore how common each category is.
Flight instructors saw cadets do worse after praise and better after a telling-off, and concluded that criticism works. Both changes were regression.
Whenever two measures are imperfectly correlated, extreme values on one go with less extreme values on the other. No cause is needed.
📝READING NOTES
Where I am in the book
Page 190-something of about 420, a month after starting. This is slower than I have read anything since university and I have stopped pretending otherwise. Parts 1 and 2 are enough for a note; if the rest changes my mind I will come back.
The two characters
| System 1 | System 2 | |
|---|---|---|
| Speed | Immediate | Slow |
| Effort | None felt | Felt, and avoided |
| Good at | Patterns, familiar situations | Rules, comparisons, arithmetic |
| Fails by | Answering confidently from too little | Not turning up |
His demonstrations are little puzzles where a wrong answer arrives first and feels right. The lesson is not distrust intuition. Trained intuition in a regular environment is excellent. The lesson is that the feeling of rightness is the same in both cases.
What you see is all there is
His phrase for System 1's habit of building a story from the available facts and never asking what else there might be. This is the mechanism under 🧩 Base-rate neglect: the base rate is not in front of me, the description is, so the description wins.
At work: one dispatcher at Kestrel says the list is too long, and I have a theory of all dispatchers by lunchtime.
Regression to the mean
The flight school story in chapter 17 is the one I retell.
- A cadet's landing is skill plus luck.
- An unusually good landing had good luck in it, so the next is probably worse. An unusually bad one, the reverse.
- The instructor praises the first and shouts at the second, and learns the wrong thing about praise.
I recognise it in three places: the carrier who gets a stern call after a terrible month and then improves, the athlete on the magazine cover, and my own best week of writing, which is always followed by an average one I then feel bad about.
Hale gets to this formally in the autumn module. For now it is filed under 00. Statistics for Product Decisions I Hub as background reading that turned out to be the main course.
My reservations
The chapter on priming rests on studies that later failed to replicate, and Kahneman has said publicly that he placed too much faith in them. I am therefore not using anything from that chapter. It does make the book an example of its own thesis: a coherent story from underpowered samples, believed by an expert.
And the two systems are so easy to picture that I catch myself treating them as places in the brain. He warns against this in chapter 1. The warning does not survive 400 pages of the metaphor.
What it produced
- 💎 A number without its base rate is a story (begun after the 1974 paper, see 📖 Tversky and Kahneman on heuristics, and strengthened here)
- 💎 Extreme results drift back toward the average on their own
Added 14 December: finished the book today, ten months and four days after starting. Parts 3 to 5 did not change this note. The second note is 📖 Kahneman on the planning fallacy.
Page Count, With Decreasing Dignity
- 14 December 2025: Finished. Nothing in parts 3 to 5 overturns this note.
- 23 March 2025: Chapters 23 and 24 read out of order for the planning fallacy note.
- 9 March 2025: Part 2 done. Note written after Castle Mound.
- 10 February 2025: Started, the day before the load test.
Reading record
- Book read: 100 %
🗣️IN SHORT
- In : Thinking, Fast and Slow, parts 1 and 2
- Kahneman argues that : most judgement is produced quickly by an automatic process that builds a coherent story from available information, and is only occasionally checked by a slower one
- by stating that : this process substitutes easy questions for hard ones, ignores base rates and sample size, and explains chance fluctuations such as regression to the mean with invented causes