Experimenter's bias
Also known as Experimenter bias
The tendency for researchers to let their expectations shape how they design a study, collect data, and interpret or report results. It overlaps with the observer-expectancy effect, and it also covers the researcher's own private decisions, such as which data to keep and when to stop.
In everyday life
A researcher who expects a new drug to work labels some disappointing results as 'measurement errors' and leaves them out.
Running an online test of two sign-up pages, you stop the test the moment your favorite pulls ahead, though you'd have let it run longer if it had been behind.
Why your mind does this
Holding steady beliefs lets us act without rethinking everything each time. Favoring information that fits what we already know is fast, protects our sense of being right, and keeps our picture of the world stable and coherent.
Its family, in 30 seconds: Confirms beliefs
How to spot it
Ask yourself whether you would have made the same call about the data if it had pointed the other way.
How to counter it
Write down your method, your stopping point, and your rules for excluding data before you see any results.