AI agents trained on sociological data sets are *whiteness* engines. Through averaging they compare inputs to an arbitrary and barely-articulable social norm. Using them fit decision making just shifts accountability for discriminatory decisions off individuals to the invisible "they".
