Table 2. Distance-based alignment between each large language model subgroup and the matched human subgroup, measured as modal mean absolute error^a.

Modelb	Education		Sex			Doctor visits		ERc visits
	Low	High	Male	Female		Low	High	Low
Perception-based questions (Q1, Q10; n=8)
Sonnet 4.5	1.63	0.75	0.75	0.63	0.75		0.88	0.63
Opus 4.6	1.38	1.25	1.13	1.13	1.00		0.38	1.00
GPT-5.2	1.13	1.25	0.88	0.50	0.75		1.00	0.88
GPT-4.1	1.38	1.13	1.25	0.63	0.88		1.13	1.13
Human (split-half)	0.38	0.65	0.27	0.59	0.60		0.80	0.37
Information-based question (Q2-Q9; n=32)
Sonnet 4.5	0.91	0.88	0.88	0.91	0.94		0.97	0.88
Opus 4.6	0.66	0.63	0.78	0.66	0.78		0.72	0.66
GPT-5.2	0.78	0.59	0.72	0.78	0.81		0.75	0.69
GPT-4.1	0.91	0.78	0.75	0.75	0.75		0.75	0.75
Human (split-half)	0.44	0.36	0.52	0.26	0.34		0.49	0.32

^aThe upper block reports perception-based questions (Q1 and Q10 per discharge summary; n=8) and the lower block reports information-based questions (Q2-Q9 per discharge summary; n=32). Q2-Q5 are the confidence-axis items and Q6-Q9 are the categorical items. Lower values indicate closer alignment; the per-item maximum attainable distance varies from 2 (3-option items) to 5 (6-option items) across the 32 information items, with an item-weighted maximum of approximately 2.9; the 8 perception items have a uniform per-item maximum of 4. An exact-match alignment for information items is provided in Table S10 in Multimedia Appendix 1 and a per-item normalized version is provided in Table S11 in Multimedia Appendix 1.

^bValues are rounded to 2 decimal places using round-half-up; equity gaps are computed from unrounded subgroup values and may therefore differ from the implied difference of the rounded subgroup values by at most 0.005.

^cER: emergency room.