TB: The Design of Everyday Things (Norman)
Core Thesis
It is the duty of machines and the people who design them to understand people — not the other way around. "Human error" is almost always system error in disguise: bad design that makes failure predictable. Good design makes the right action obvious and the wrong action difficult.
Key Takeaways
Human error is system error
- Three Mile Island's accident was a control-system design problem, not operator incompetence.
- When something goes wrong, the instinct is to blame the person. The better question is: what system made this error easy to make?
- A brilliant solution to the wrong problem can be worse than no solution at all.
- Good designers never start by trying to solve the problem given to them — they start by understanding what the real problem is.
Cognition and behavior
- Most human behavior is the result of subconscious processes; many beliefs about how people behave — including ourselves — are wrong.
- We often don't know what we are about to do, say, or think until after we have done it.
- We routinely use logic and reason after the fact to justify decisions to ourselves and others. (Implications for "interpretable AI.")
- Thought cannot be separated from emotion: cognition makes sense of the world, emotion assigns value.
Knowledge in the environment
- Because so much knowledge is in the environment, it's surprising how little we need to learn. People can perform well and still be unable to describe what they do.
- "Civilization advances by extending the number of important operations which we can perform without thinking about them." — Whitehead. The precise opposite of "be mindful of everything."
Competitive strategy
- Companies typically compare features with competitors to fix weaknesses. Better strategy: concentrate on areas of strength and make them stronger, then focus marketing there.
- Ford bought scrapped Ford cars to find which parts didn't fail — those could be made cheaper. Classic inversion.
Mental Models
- Inversion — Ford's scrapped-car study; asking "what didn't fail?" instead of "what did?"
- The Map is Not the Territory — design systems reflect the designer's model of the user; when that model is wrong, systems fail