Experimenter’s bias is a form of bias in which a researcher’s expectations or assumptions influence the design, interpretation, or reporting of an experiment. In AI evaluation, it can affect dataset choices, metric interpretation, and conclusions about model behavior.
See confirmation bias.