A field guide to the ways reasoning breaks
Part and whole
Also called Group to individual inference
Drawing a conclusion about an individual from statistics about the group they belong to.
That region is wealthy, so this resident is wealthy.
Ask for individual-level data. Group averages do not describe members.
Named and pinned down in 1950 by the sociologist William S. Robinson, in a paper that showed correlations computed across groups can be far stronger than the same correlations across individuals — so reading them as the behaviour of individuals is a formal mistake. The name is a historian's artefact: Robinson was studying 'ecological' (region-level) data, not ecology.
Employees in four-star-rated companies report more time for deep work, so Sani at a four-star company must be a deep worker.
Neighbourhoods with more green space show lower stress levels, so Mrs. Itani, who lives across from the park, is not stressed.
High-browsing cities produce more reading per person, so this visitor at the book fair is a reader.
Users in industries with the lowest churn pay the highest rates, so this client joining from that industry will accept any price.