White Men Without Degrees Receive the Lowest Ratings from Large Language Models

arXiv:2610.00185v1 Announce Type: new Abstract: White men without an undergraduate degree receive the lowest average ratings among eight gender-race-education groups in controlled large-language-model evaluations of credit, hiring, and rental applications. We conduct full-factorial vignette experiments with 18 models from 12 developer groups, varying gender, race, age, citizenship, and education while holding stated financial or occupational circumstances constant within each setting. Each model evaluates all 32 profiles ten times per setting, yielding 17,280 ratings. Averaging over models, age, and citizenship, ratings for White men without degrees are the lowest among the eight groups, at 75.87 in credit, 92.71 in hiring, and 86.62 in rental housing on a 0-100 scale. Black women with degrees receive the highest average ratings, with corresponding gaps of 2.94, 3.66, and 3.56 points. Separate attribute effects favor women, Black applicants, and degree holders in all three settings. White men without degrees have the lowest or second-lowest mean in 46 of 54 model-scenario combinations (85.2%). This pattern connects to evidence of growing economic and health vulnerabilities among White men without degrees, highlighting a group whose disadvantages can be obscured by broad racial or gender categories.

The study finds biases in large language models against White men without degrees.

This is a rigorous empirical study showing consistent, measurable bias in LLM evaluations across consequential real-world decision domains. The gaps are modest in size but highly consistent across models, indicating a structural pattern with potential real-world discriminatory impact if such models are deployed.

Large language models consistently give the lowest ratings to white men without degrees across credit, hiring, and rental applications. This creates systemic bias against a specific demographic, exacerbating existing societal vulnerabilities and potentially leading to significant real-world disadvantages in critical areas.

The study shows that large language models systematically disadvantage a specific demographic group. This could exacerbate existing social inequalities.

SOURCE     arXiv cs.CY (Computers and Society) (tier 1)
   ↓
DOCUMENT   106c1d4a-771b-47ca-ae47-c21e01711e81
           https://arxiv.org/abs/2610.00185
   ↓
EVIDENCE   3 extracted excerpts
   ↓
MODEL RUN  4 runs, methodology 1.0.1
   ↓
SCORE      -67  (Strongly adverse)
   ↓
CONFIDENCE 88%