CS Paper Radar

Name-Bias in Large Language Models

Kalyan Bhetwal, Ceren Mert, Ian Whitehouse

cs.CL ·submitted 2026-10-15 15:19 UTC · v1

Abstract

Names carry demographic indicators such as age, gender, and race, which makes them a subtle but strong target of bias in large language models (LLMs). In this work, we present an evaluation of name-based bias across 6 local large language models spanning multiple providers and parameter scales. Using a fill-in-the-blank task, we prompt these models to choose between pairs of first names drawn from U.S. Social Security Administration datasets and a race classification dataset by Rosenman et al.. Our experimental de- sign isolates age and gender effects through balanced name pairings and sentence templates that share a fixed syntactic structure. Across 300,000 local LLM responses, we find consistent and measurable name bias. Models prefer names that are less stereotypically male, older, popular, and white. Furthermore, we find that name-bias is highly sentence-dependent, but not more than associated stereotypes, and gender is the most significant bias, followed by age, name popularity, and race. Our results demonstrate that name-based bias remains pervasive even in local LLMs, highlighting the necessity for standardized evaluation methods and mitigation strategies.

Closest papers

Nearest neighbours by meaning across the whole index.

Who Gets a Token, and What Does It Carry? Unequal Name Support and Concept Access in Large Language Models

cs.CL cs.AIcs.CYcs.ET+1 ·2026-09-28 ·Mir Tafseer Nayeem, Davood Rafiei
similarity 0.73
Abstract · Preprint

Names are personal identifiers, but they also carry social meaning and are widely used to evaluate how language models treat different people. Such evaluations typically assume that matched names are comparable model inputs. We show that this assumption often fails at the lexical interface: matched names are not necessarily matched inputs. Some names receive direct single-token access, while others are assembled from multiple subwords, creating unequal name-surface support. Across nearly half a million first names and 12 LLM-associated tokenizers, direct lexical access is highly selective, model dependent, and uneven across race- and gender-associated name metadata. We introduce NameTrace, a model-native, fine-grained, pre-behavioral framework for measuring whether unequal name-surface support remains a vocabulary property or becomes visible in task-relevant internal representations. NameTrace measures concept accessibility from the model's own probabilities over task-specific adjective axes with continuous task-aligned weights. On matched atomic and short-fragmented names within the same race/ethnicity--gender-associated strata, support predicts systematic differences in concept accessibility across fellowship, hiring, clinical assessment, and lending. These differences persist across all eight matched strata, extend across model families, and transfer to unseen names. Hidden-state interventions further show that the measured task directions have downstream leverage, shifting later constrained choices. Unequal lexical support is therefore demographically structured at the input and remains visible in task-relevant model computation. NameTrace makes lexical comparability measurable, supporting a broader principle: behavioral comparability begins with lexical comparability.

0.33

Who Gets Counted? Large Language Models, Informative Refusal, and Downstream Bias in Name-Based Demographic Classification

cs.CY ·2026-11-04 ·Christian Pipal, Morgan Wack, Eva-Maria Vogel
similarity 0.72
Abstract

Researchers routinely infer gender, race, and ethnicity from names, yet the standard tools used for this task misclassify the groups that diversity research most needs to measure. Large language models (LLMs) have been proposed as a remedy because they label diverse names accurately. We ask whether that accuracy carries over to the estimates researchers report. Across 10,000 Chinese names, 19,998 North Carolina voters with self-reported race, and 300 political science authors with human reference codes, we compare three LLMs with seven traditional tools and link the voter sample to party registration and age. Selective labeling explains part of the LLMs' accuracy advantage. Claude Haiku labels half of the voters, and a traditional tool limited to its most confident half is more accurate. Hard labels from every tool understate the White--Black gap in Democratic registration by 36% to 65%, while BIRDiE estimation on BISG posteriors comes within 0.03 of it. Haiku's missing labels select on age, inflating the age gap by six years. Combined with BISG, LLM answers sharpen individual labels without a demonstrated gain in BIRDiE accuracy. Classifiers should be judged by the estimates they produce, including the consequences of missing labels

0.19

Pre-Trained Language Models as Zero-Shot Tools for Social Science Research

cs.CL ·2026-10-13 ·Noah Dasanaike
similarity 0.69
Abstract

Record linkage and demographic inference from names are routine demands in social science. Most approaches to these problems require labeled training data that are difficult or costly to acquire. I show that pre-trained language models perform both tasks accurately without requiring task-specific training. For record linkage, I introduce EnsembleLink, an algorithm that uses pre-trained embeddings alongside string-based scorers to retrieve a short list of candidates and then identify the likeliest matches. EnsembleLink matches or exceeds the performance of most prior approaches while generalizing across scripts and formats that those methods are unable to handle. For ethnicity inference from names, I show that large language models (LLMs) achieve up to 83.8% accuracy on U.S. voter files, compared to 78.0% for name-supplemented BISG. Open-weight models offer reproducible implementations that approach these models in accuracy. Moreover, LLMs are able to infer group membership for religious, caste-based, and other ethnic group categories outside of the United States, and reduce bias in downstream estimates of geographic racial gaps. Where pre-training encodes the world knowledge required by each task, task-specific labels add less information.

0.33

From Language to Country to Valence: A Multilingual Audit of Stereotype Bias Across LLMs

cs.CL ·2026-11-02 ·Zihao Sun
similarity 0.69
Abstract · 10 pages. Author preprint; conference review ongoing

Large language models (LLMs) are increasingly considered for education, recruitment, workplace decision support, and other consequential applications, making pre-deployment social-bias detection essential. Yet major benchmarks are English-centred or translated, focus mainly on negative stereotypes, and may miss linguistic and cultural variation. We investigate gender, racial, and religious bias using 12,000 locally written Reddit posts in English, Spanish, French, and German. Only search terms are translated, preserving local context. The dataset is balanced by language and category; English gender is additionally balanced by stereotype valence. Three open-weight and three proprietary API-based models are evaluated with a two-blank design that reduces grammatical-gender cues. Religion has the highest absolute choice imbalance in every model-language condition. Spanish is highest for gender, whereas German is highest for race and religion. The evaluated API models score below the open-weight models throughout. Negative English-gender items exceed positive items; Australian and Mexican contexts exceed their Canadian and Spanish counterparts. These descriptive differences measure choice concentration, not stereotype endorsement or causal country effects. The crossed patterns show how aggregate scores can conceal category- and context-specific behaviour, motivating closer inspection before deployment.

0.15

It's How You Ask: Gender-Associated Linguistic Bias in LLMs

cs.CL cs.AI ·2026-08-13 ·Katherine Van Koevering, Anjalie Field
similarity 0.66
Abstract

Professional communication is increasingly mediated by LLMs - but do these models serve all users equally? We show that when prompts contain linguistic features more commonly used by women (hedges, tag questions, collective reference), they systematically elicit shorter, less sophisticated, and less formal responses across three document types and four models. These effects persist after controlling for prompt complexity and feature carry-over. Explicit gender cues like sign-off names are encoded in the same representational space as linguistic dialect - suggesting shared underlying mechanisms - yet linguistic register is far more influential, producing large, consistent effects where names produce none. Our results further reveal that post-hoc mitigation is challenging: because these patterns are culturally embedded and outside conscious control, users cannot easily avoid them through strategic self-presentation, and mechanistic analysis reveals that linguistic features are encoded in early transformer layers and entangled with other features. Our work calls for upstream consideration of the influences of linguistic variation to mitigate disparate impacts of LLM-mediated workplace communication.

0.04

Revealing the Biases of Large Language Models By Inferring Prior Distributions

cs.LG ·2026-11-13 ·Joshua Ryo Cheng, Thomas L. Griffiths, Jake C. Snell
benchmark similarity 0.65
Abstract

Large Language Models (LLMs) have been known to exhibit a wide range of social biases including name-nationality associations, occupation-gender associations and racial stereotypes. Existing approaches to evaluating these biases typically involve measuring surface-level behavior where a model is fed a dataset of prompts and its outputs are scored relative to ground-truth answers. Inspired by prior elicitation in cognitive science, we propose a different perspective that reconstructs latent bias as a prior underlying the model's behavior. We formulate the problem as a Bayesian inverse problem where, given prompt-response observations on a bias benchmark dataset, we infer a low-dimensional latent distribution that summarizes the model's stereotyping tendencies. This paper develops the mathematical foundation, a practical reconstruction algorithm, and experiments using the Bias Benchmark for QA (BBQ) benchmark to infer the prior distributions of LLMs over responses in ambiguous social contexts.

0.11

LLMs Cannot Be Neutral: Gender Sensitivity Across Candidates, Evaluators, and Models

cs.CL cs.HC ·2026-11-02 ·Danial Amin, Joni Salminen, Bernard J. Jansen
hot direction similarity 0.65
Abstract · This manuscript is under review at FAccT'2027

Large language models (LLMs) are increasingly used to evaluate candidates for professional roles, raising concerns about their sensitivity to demographic identities. We examine how candidate and evaluator gender assignments affect decisions by LLM-as-a-Judge (LaJ) models and decision language models (DLMs). In a hypothetical city council voting task, eight models evaluated 240 synthetic professional narratives from NVIDIA's US and French Nemotron-Personas datasets under independently varied gender assignments, producing 30,720 judgments. Explicit candidate gender labels increased Yes-vote probability by 1.05-1.73 percentage points relative to unspecified gender, while only the nonbinary evaluator assignment showed a significant increase (1.26 points). Identity matching showed no significant overall association. Yes-vote rates varied across models (81.3-97.2%), and 2.5-3.3% of decisions reversed relative to the unspecified-identity baseline. Under direct substitutions between explicit gender identities, DLMs exhibited fewer decision reversals than LaJ models (1.18% vs. 3.38%), indicating lower observed decision-level gender sensitivity among the tested DLMs. These findings demonstrate that gender sensitivity depends on candidate and evaluator identities and the evaluation model, while highlighting the importance of examining individual decision consistency alongside aggregate differences.

0.31

Linguistic Triggers of Gender and Racial Bias in Open-Weight LLMs Applied to Recruitment

cs.CL cs.AIcs.CY ·2026-09-16 ·Kosuke Kitahara, Nobuhiro Yamaguchi
published similarity 0.65
Abstract · Accepted at the 9th AAAI/ACM Conference on AI, Ethics, and Society (AIES 2026). Extended version with Appendices A-B (prompt templates and f…

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.

0.35