Table 1.

Correspondence between confidence classification methods in the real dataset.

Classification Method #1	Measure	Classification Method #2
		Human Coder 1	Human Coder 2	AI Model	Trichotomized
Human Coder 1	Correlation (r)	-
	% agreement	-
	Cohen’s k	-
Human Coder 2	Correlation (r)	0.748	-
	% agreement	81.3%	-
	Cohen’s k	0.648	-
AI Model	Correlation (r)	0.381	0.504	-
	% agreement	34.7%	26.7%	-
	Cohen’s k	0.207	0.250	-
Trichotomized Numeric Confidence	Correlation (r)	0.689	0.708	0.337	-
	% agreement	46.7%	42.7%	28.0%	-
	Cohen’s k	0.515	0.506	0.164	-

Note. N = 75.