Overfitting
Etymology
A statistics and machine-learning term: a model "fits" the data it was trained on, and the "over-" marks fitting carried to excess, so it learns the sample’s noise rather than the underlying signal. Beyond computing, it names any conclusion contoured too closely to a handful of cases. The term is descriptive.
noun · /OH-ver-fit-ing/
Tuning a model or belief so tightly to past data that it captures the noise and quirks instead of the real pattern, and fails on anything new.
What it names
A handle for drawing too specific a lesson from too little experience — the strategy that "worked last time" but only because it memorized the accidents of one case rather than a general truth.
Example
Their hiring rubric was overfitting to their two star employees, screening out everyone who didn’t share those two people’s hobbies.
Related idea
The systematic cousin of apophenia: apophenia is the human tendency to see patterns in noise, and overfitting is what happens when a model or method does exactly that as a rule.