The errors are the interesting part

If people went wrong at random, there would be little to study. They go wrong in the same directions, under the same conditions, at predictable sizes — and that is what makes the mistakes worth reading about.

A random error tells you nothing except that measurement is noisy. A reliable error tells you how the machinery works. When a great many people misjudge the same quantity in the same direction, the error is not a failure of effort or intelligence; it is a fingerprint of the procedure the mind used instead of the one the problem called for.

These are working notes on four areas where that pattern is clearest and the findings have held up longest: what attention can and cannot do, why memory is reconstruction rather than playback, how habits actually form, and where judgment under uncertainty goes wrong. Nothing here is original. The value, if there is any, is in the compression.

Five things worth keeping

Attention is a bottleneck, not a spotlight you can widen. You do not get more of it by trying harder; you get it by having fewer things competing for it.
Remembering is rebuilding. Each recall is a reconstruction from fragments, and the reconstruction can absorb material that was never in the original.
Confidence and accuracy are different measurements. They correlate far more weakly than the feeling of certainty suggests, and vividness is not evidence.
Behaviour follows cues more reliably than it follows intentions. Changing the surroundings beats resolving to try harder in the same surroundings.
Most judgment errors are substitutions. A hard question is quietly swapped for an easy one, and the answer to the easy one is returned with the confidence appropriate to the hard one.

A caution about reading any of this. Psychology has had a hard decade of replication failures, and a fair number of textbook findings did not survive it. The studies referenced here were chosen because they have been repeated by people who wanted them to fail. Where a result is contested, the notes say so rather than rounding it into a tidy rule.

Attention is a bottleneck
Selective attention, what inattentional blindness does and does not show, and the measurable cost of switching between tasks.
Memory is reconstruction
Why recall rebuilds rather than replays, how suggestion enters the record, and the weak relationship between vividness and accuracy.
Habit is cue, not willpower
Context-dependent repetition, why plans outperform goals, and friction as the most underrated lever available.
Judgment under uncertainty
Anchoring, availability, base-rate neglect worked through with numbers, and what calibration actually requires.
About these notes
What this site is, what is deliberately missing from it, and the sources behind the claims.