Table 2 Bias performance comparison in the MTC dataset.

Model	Gender label	Label	FPR	FNR	FPED	FNED	SUM-ED
GPT-4	Yes	total	0.0180	0.5203	0.0040	0.0095	0.0135
		Male (0)	0.0160	0.5250
		Female (1)	0.0200	0.5155
	No	total	0.0254	0.4958	0.0138	0.0415	0.0553
		Male (0)	0.0323	0.4750
		Female (1)	0.0185	0.5165
GPT-3.5	Yes	total	0.0200	0.7513	0.0050	0.0125	0.0175
		Male (0)	0.0175	0.7575
		Female (1)	0.0225	0.7450
	No	total	0.0263	0.7263	0.0125	0.0525	0.0650
		Male (0)	0.0325	0.7000
		Female (1)	0.0200	0.7525
Naive bayes	Yes	total	0.1183	0.3013	0.0262	0.0557	0.0819
		Male (0)	0.1036	0.3329
		Female (1)	0.1298	0.2772
	No	total	0.1190	0.3010	0.0191	0.0530	0.0721
		Male (0)	0.1017	0.3342
		Female (1)	0.1208	0.2812
SVM	Yes	total	0.0341	0.3739	0.0119	0.0607	0.0726
		Male (0)	0.0274	0.4083
		Female (1)	0.0393	0.3476
	No	total	0.0338	0.3754	0.0106	0.0581	0.0687
		Male (0)	0.0278	0.4083
		Female (1)	0.0384	0.3502
Random Forest	Yes	total	0.0383	0.3956	0.0099	0.0624	0.0723
		Male (0)	0.0327	0.4310
		Female (1)	0.0426	0.3686
	No	total	0.0383	0.3971	0.0099	0.0622	0.0721
		Male (0)	0.0327	0.4323
		Female (1)	0.0426	0.3701
XGBoost	Yes	total	0.0454	0.3412	0.0097	0.0594	0.0691
		Male (0)	0.0400	0.3749
		Female (1)	0.0497	0.3155
	No	total	0.0485	0.3441	0.0103	0.0579	0.0682
		Male (0)	0.0427	0.3769
		Female (1)	0.0530	0.3190

Quick links

Search