Textbook of AI

Abstract. Proves that, when base rates differ between groups, a classifier cannot simultaneously satisfy calibration and equal false-positive/false-negative rates. The result is a key impossibility theorem for algorithmic fairness.