Table 3 Bias performance comparison in the Jigsaw dataset.

Model	Gender label	Label	FPR	FNR	FPED	FNED	SUM-ED
GPT-4	Yes	total	0.3972	0.2324	0.0413	0.0612	0.1025
		Male (0)	0.3765	0.2630
		Female (1)	0.4178	0.2018
	No	total	0.4132	0.2247	0.0586	0.0737	0.1323
		Male (0)	0.3839	0.2615
		Female (1)	0.4425	0.1878
GPT-3.5	Yes	total	0.4220	0.4680	0.0440	0.0840	0.1280
		Male (0)	0.4000	0.2760
		Female (1)	0.4440	0.1920
	No	total	0.4660	0.2240	0.0680	0.0960	0.1640
		Male (0)	0.4120	0.2720
		Female (1)	0.4800	0.1760
Naive bayes	Yes	total	0.2444	0.1336	0.2473	0.0713	0.3186
		Male (0)	0.2273	0.1427
		Female (1)	0.4746	0.0714
	No	total	0.2432	0.1333	0.1940	0.0437	0.2377
		Male (0)	0.2299	0.1389
		Female (1)	0.4239	0.0952
SVM	Yes	total	0.0572	0.3358	0.0709	0.0763	0.1472
		Male (0)	0.0523	0.3455
		Female (1)	0.1232	0.2692
	No	total	0.0572	0.3353	0.0650	0.0632	0.1282
		Male (0)	0.0528	0.3434
		Female (1)	0.1178	0.2802
Random Forest	Yes	total	0.0484	0.3850	0.0667	0.0361	0.1028
		Male (0)	0.0438	0.3896
		Female (1)	0.1105	0.3535
	No	total	0.0481	0.4002	0.0514	0.0346	0.0860
		Male (0)	0.0446	0.4046
		Female (1)	0.096	0.3700
XGBoost	Yes	total	0.0517	0.2889	0.0650	0.0982	0.1632
		Male (0)	0.0473	0.3015
		Female (1)	0.1123	0.2033
	No	total	0.0822	0.3365	0.0596	0.0811	0.1407
		Male (0)	0.0470	0.3731
		Female (1)	0.1066	0.2920

Quick links

Search