When Algorithms Favor the Underrepresented – Race and Gender Biases in LLM Résumé Evaluations

S. Kim et al.

Journal of Personnel Psychology2026https://doi.org/10.1027/1866-5888/a000388article
AJG 2ABDC B
Weight
0.50

Abstract

Abstract: The literature on name-based biases in hiring suggests pervasive discrimination, as White-sounding names receive more callbacks than Black-sounding ones. This study assesses whether AI systems, through open Large Language Models (LLMs), exhibit similar biases. The LLMs evaluated résumés on attributes such as competence and warmth, aggregating these dimensions into composite scores for each résumé. The different names attached to a résumé led to changes in evaluation, despite identical content. Statistically significant race and gender biases were found in most models for warmth and competence ratings. Unlike typical settings, Black applicants and female names were rated slightly higher through the LLMs’ evaluations. These findings highlight the importance of examining AI tools used in hiring as they may unintentionally reflect societal biases.

Open via your library →

Cite this paper

https://doi.org/https://doi.org/10.1027/1866-5888/a000388

Or copy a formatted citation

@article{s.2026,
  title        = {{When Algorithms Favor the Underrepresented – Race and Gender Biases in LLM Résumé Evaluations}},
  author       = {S. Kim et al.},
  journal      = {Journal of Personnel Psychology},
  year         = {2026},
  doi          = {https://doi.org/https://doi.org/10.1027/1866-5888/a000388},
}

Paste directly into BibTeX, Zotero, or your reference manager.

Flag this paper

When Algorithms Favor the Underrepresented – Race and Gender Biases in LLM Résumé Evaluations

Flags are reviewed by the Arbiter methodology team within 5 business days.


Evidence weight

0.50

Balanced mode · F 0.40 / M 0.15 / V 0.05 / R 0.40

F · citation impact0.50 × 0.4 = 0.20
M · momentum0.50 × 0.15 = 0.07
V · venue signal0.50 × 0.05 = 0.03
R · text relevance †0.50 × 0.4 = 0.20

† Text relevance is estimated at 0.50 on the detail page — for your query’s actual relevance score, open this paper from a search result.