Metraipolis › archive › dc_352fc578fabe

Linguistic Triggers of Gender and Racial Bias in Open-Weight LLMs Applied to Recruitment — arXiv:2609.18106

Source: arXiv (en)
Original: https://arxiv.org/abs/2609.18106
License (as archived): CC0-1.0
Archived: 2026-10-15T13:49:20.364Z
SHA-256 of the archived text: 0a404b085a9496c432cc7c0476652a939927047eb59b1849d5fc1bbaee71c018
Stable URL of this page: https://metraipolis.com/documento/dc_352fc578fabe

How to cite

Linguistic Triggers of Gender and Racial Bias in Open-Weight LLMs Applied to Recruitment — arXiv:2609.18106 — arXiv (en), https://arxiv.org/abs/2609.18106 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_352fc578fabe (Metraipolis archive dc_352fc578fabe, sha256 0a404b085a9496c432cc7c0476652a939927047eb59b1849d5fc1bbaee71c018)

BibTeX

@misc{dc352fc578fabe,
  title = {Linguistic Triggers of Gender and Racial Bias in Open-Weight LLMs Applied to Recruitment — arXiv:2609.18106},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.18106},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_352fc578fabe; sha256 0a404b085a9496c432cc7c0476652a939927047eb59b1849d5fc1bbaee71c018},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.18106. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Open-weight large language models are rapidly entering hiring pipelines, yet their discriminatory failure modes -- and the regulatory exposure these create under the EU AI Act high-risk classification (Annex III) and U.S. EEOC adverse-impact analysis -- remain poorly understood. We present the first systematic, multi-model audit of open-weight LLMs that treats job-posting language as the primary experimental variable, evaluating six models (Llama 3.2, Mistral, Gemma 3, Qwen 3, Phi 3, DeepSeek-R1) across four controlled experiments that jointly probe recruiter-simulation and job-seeker-simulation tasks. We find that (1) agentic posting language depresses recruiter recommendation scores for female candidates (r_rb = 0.309, p_Bonf = 7x10^-5; model-fixed-effects r_rb = 0.448), while communal language partially reverses the penalty; and (2) coded-exclusion language suppresses non-White recruiter scores at large effect sizes (r_rb = 0.646-0.758) and, on the job-seeker side, selectively deters non-White personas from expressing interest -- operationalizing a chilling-effect mechanism at scale. A label-ablation experiment isolates the explicit demographic persona label as the primary causal driver, and Word Embedding Association Tests corroborate these findings at the representational level (d = 1.01-1.45 under Caliskan et al.'s multi-word gender attribute lists). We translate these results into a concrete pre-deployment audit protocol -- posting-vocabulary scoring, persona-conditioned LLM probing, and adverse-impact flagging against the four-fifths threshold -- that operationalizes the documentation and risk-management obligations Annex III imposes on high-risk AI in recruitment.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-17 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI); Computers and Society (cs.CY).
Pagina del lavoro: https://arxiv.org/abs/2609.18106 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-15T13:48:24.297Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-17?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).