A field guide to the ways reasoning breaks
Evidence and sampling
Also called Linda problem
Judging a detailed, specific scenario as more likely than the broader one that contains it.
She is more likely a banker who volunteers than just a banker.
Remember that adding detail can only lower probability, never raise it.
Documented by Amos Tversky and Daniel Kahneman in 1983 through the 'Linda problem': told that Linda is bright and outspoken, most people rated 'Linda is a bank teller and feminist' as more likely than 'Linda is a bank teller'. The finding held even with statistics students, which is what made it famous.
A jury thinks it more likely the defendant 'planned the fraud and covered it up' than just that he 'planned the fraud' — the second story is more damning but strictly less probable.
An investor judges 'the startup will succeed because it rides the AI wave with a brilliant team' as more plausible than 'the startup will succeed'.
A football commentator says the striker is 'more likely to score and celebrate toward the ultras' than just 'more likely to score'.
A dating profile reader concludes a match is 'more likely a surgeon who rescues stray cats' than just 'a surgeon'.