Berkson's Paradox

mental-model

When you compare only cases that passed the same selection threshold, that filter can create an apparent trade-off between qualities that are not opposed in the wider population.

A dating app can appear to offer a grim choice: attractive people who are jerks, or less attractive people who are smart. But that pattern may describe neither people nor dating—it may describe who survived the app’s filter.

E1

The missing quadrants create the correlation

Suppose several qualities can help someone enter a visible pool. A person strong on one quality may pass despite being weak on another; someone weaker on the first may need more of the second to pass. Once you inspect only those who cleared the threshold, the surviving cases line up as if the qualities trade off. The filter has removed the combinations that would expose the illusion. Berkson’s Paradox is therefore a specific form of Selection Bias: conditioning on admission creates a relationship inside the sample that need not exist outside it.

E1

Where it shows up

The dating-app shortlist

If the pool disproportionately leaves you comparing “good-looking jerks” with “okay-looking smart people,” the absent quadrants matter most. Their exclusion makes attractiveness and intelligence or kindness look negatively linked among the people you see.

E1

Not every ugly trade-off is artificial

The paradox does not prove that the suspected relationship is false. It says the selected pool cannot establish it by itself: a real trade-off and a filter-created trade-off can look alike unless you examine who never entered the sample.

E1

Reconstruct the rejected quadrants

Tomorrow, when a shortlist seems to reveal that one desirable quality comes at the cost of another, draw a two-by-two grid of both qualities. Identify which cells your admission rule makes hard to observe, then withhold the stereotype until you can compare against people outside the selected pool.

E1

Episodes that teach this