"Human Error? No, Bad Design"
The book's second great thesis (alongside the design principles): what we call "human error" is usually the result of poor design; it should be called system error / design error.
Core claims:
- Humans err continually — it is intrinsic to our nature. A system where a single act by a single person can cause calamity is badly designed. "If the system lets you make the error, it is badly designed. If it induces you to make the error, it is really badly designed."
- We should treat human failures like mechanical failures: find the fundamental cause and redesign so it can't recur (or has minimal impact). Instead, society does "blame and punish; blame and train" — which feels good but fixes nothing; the same error recurs.
- People wrongly blame themselves ("I'm bad at mechanical things") even for design faults, partly because society/the legal system embed the "person at fault" idea. The Return-vs-Enter keyboard story: everyone made the error but reported nothing because they blamed themselves.
- Designs fail because they serve the requirements of machines (precision, constant alertness, no interruptions) over the requirements of people (creative, flexible, interruptible, bad at precision). Major error drivers: interruptions, time stress, and the attitude toward error.
Why people err (root drivers)
Tasks demanding unnatural behavior (hours of alertness, precise input, multitasking) + interruptions + the blame attitude. Three Mile Island: "human error" was the immediate verdict, but the control rooms were so poorly designed error was inevitable.
Credo (ch.2)
Eliminate the term "human error." Treat human–machine interaction as collaboration / communication; provide help and guidance instead of error messages; assume actions are partially correct; never make people start over; think positively. Failure = a learning experience ("Fail often, fail fast").
Caveat
Sometimes people really are at fault — sleep deprivation, drink, drugs, deliberate violations (routine/situational). But the far greater share of accidents is poor design.
Confidence
1.0 — repeated, explicit thesis.
Sources: ["DOET ch.5 p.162-169; ch.2 'Falsely Blaming Yourself' p.65-67"]