There is a long history of various kinds of "information inbreeding" in AI and data analysis: the tendency of some inexperienced researchers to overfit their re

There is a long history of various kinds of "information inbreeding" in AI and data analysis: the tendency of some inexperienced researchers to overfit their regression models, using bootstrapping methods to expand small data samples, etc. Nothing new here. Because of laziness human beings just love a Confirmation bias. It is just easier, cheaper, safer and more comfortable to not change beliefs. Without control, AI will reflect that.