The practice, by an algorithmic model, of producing differential outcomes by geography that correlate with the racial composition of neighborhoods, typically through facially neutral geographic or neighborhood-characteristic variables that encode the effects of historical redlining.