Heuristics and biases

The research programme Kahneman and Amos Tversky founded: people judge under uncertainty using heuristics — fast, serviceable shortcuts — and those shortcuts produce biases, errors that are systematic rather than random. The systematic part is what made it a science. A random error is noise; a predictable one is a mechanism you can study, and eventually a mechanism you can design around.

The focusing-illusion is this spoke’s worked example: attention inflates importance, reliably, in a direction you can state in advance.

The dunning-kruger-effect has the same shape on paper: a systematic error in self-assessment, stated as a direction, and the programme’s most famous export to the general public. It is also the place where “systematic” turns out to be doing less work than it looks like. A censored-regression model carrying noise and a random floor and ceiling, and no psychological parameter, fits real self-predictions almost perfectly (magnus-peresetsky-2022). Systematic error is evidence of a mechanism only when noise wouldn’t have produced it — though in this case the original authors ran an experiment that noise does not explain (kruger-dunning-1999), and the argument is still live.

System 1 and System 2

Kahneman’s popular framing splits judgment into a fast, automatic, effortless mode and a slow, deliberate, effortful one. The fast system is not a defect — it handles nearly everything, and handles it well — but it answers a slightly easier question than the one asked, and the slow system usually ratifies rather than checks.

That vocabulary has travelled far beyond psychology, and this hub is a case in point: ../defensive-security-wiki’s soc-fast-slow-thinking designs a two-tier security-operations centre around it, machine triage as System 1 and analysts as System 2. Useful metaphor, and worth naming as a metaphor: an alert pipeline and a mind are not the same object, and the framing’s authority in the borrowed setting comes from the psychology rather than from evidence about pipelines.

Two selves, and why the happiness studies disagree

thinking-fast-and-slow adds the piece that makes the focusing-illusion‘s evidence cohere: the experiencing self who lives through an episode and the remembering self who scores it afterwards. The remembering self works by the peak-end rule — the most intense moment and the ending — and largely ignores duration.

So the income-and-happiness split isn’t two measurements of one thing disagreeing. Asking someone how satisfied they are queries the remembering self; sampling their moments queries the other one. Both have a claim on the word “happiness,” and which one a study measures decides its answer.

The standing caveat

This is the sub-field the replication crisis hit hardest in the public imagination, priming most notoriously. The damage reaches the book itself: its Chapter 4 citations carry a replicability index of 14, and Kahneman conceded he “placed too much faith in underpowered studies.” The core judgment-under-uncertainty work has held up better than the social-priming literature around it, but “Kahneman said so” is not evidence — see replication-crisis, and this spoke’s CLAUDE.md rule to record the study behind a claim.

Related

focusing-illusion · dunning-kruger-effect · daniel-kahneman · thinking-fast-and-slow · replication-crisis · kahneman-focusing-illusion · synthesis