confirmation bias

Terms from Statistics for HCI: Making Sense of Quantitative Data

Confirmation bias is when we unintentionally only seek out evidence or take notice of results that reinforce our existing views. With traditional statistics this might lead us not to publish results that we think are wrong, perhaps even removing data points that we feel must have been errors of some kind. In Bayesian statistics confirmation bias can cause even more problems, as the choice of prior probability is effectively one's own beliefs encoding (belief) as probability and hence the output of the process, the posterior probability, may simply reflect one's existing prejudices.

Used on pages 60, 69, 77, 79