Table 1.

Accuracy and fairness (with respect to demographic parity) for various methods on the balanced test set of the Adult dataset.

Algorithm	Fair → 1.0←	Accuracy ↑	Fair → 1.0←	Accuracy ↑
GP	0.80 ± 0.07	0.888 ± 0.007	0.54 ± 0.05	0.900 ± 0.006
LR	0.83 ± 0.06	0.884 ± 0.007	0.52 ± 0.03	0.898 ± 0.003
SVM	0.89 ± 0.06	0.899 ± 0.004	0.49 ± 0.05	0.913 ± 0.004
FairGP (ours)	0.86 ± 0.07	0.888 ± 0.006	0.87 ± 0.09	0.902 ± 0.007
FairLR (ours)	0.90 ± 0.06	0.874 ± 0.009	0.93 ± 0.04	0.886 ± 0.012
ZafarAccuracy (Zafar et al., 2017b)	0.67 ± 0.17	0.808 ± 0.016	0.77 ± 0.08	0.853 ± 0.017
ZafarFairness (Zafar et al., 2017b)	0.81 ± 0.06	0.879 ± 0.009	0.74 ± 0.11	0.897 ± 0.004
Kamiran and Calders (2012)	0.87 ± 0.07	0.882 ± 0.007	0.96 ± 0.03	0.900 ± 0.004
Agarwal et al. (2018)	0.86 ± 0.08	0.883 ± 0.008	0.65 ± 0.04	0.900 ± 0.004

Fairness is defined as PR_{s = 0}/PR_{s = 1} (a completely fair model would achieve a value of 1.0). Left: using race as the sensitive attribute. Right: using gender as the sensitive attribute. The mean and std of 10 repeated experiments.