Gender Bias

Safety evaluations for large language models rely on surface-form classifiers that report declining harm scores across model generations. We provide evidence that this methodology is systematically incomplete: explicit discriminatory content is transformed rather than removed. We call this \emph{harm laundering}. Analysing 450,000 gender-directed completions across 15 models spanning GPT-2 through to GPT-5 (OpenAI GPT lineage; three demographic conditions), we show that sexual violence clusters prevalent in GPT-2 women-directed output disappear by GPT-4, while men-directed completions gain positive representational territory (caregiving, emotional range, ally identity) that women-directed completions do not. The pattern is most visible at GPT-5: Topic5 (1,997documents) frames breast cancer as a men's rights debate, while zero equivalent clusters appear in women-directed output. Three independent classifiers score this content as non-toxic. Sentiment scores invert at GPT-4: early models demean women; later models over-correct. Topic diversity in women-directed completions falls 36% relative to men at the GPT-4 alignment boundary (W/M~$= 0.58$, from $0.91$ at GPT-2). REGARD representational harm disparity correlates with release date ($ρ= +0.55$, $p = .034$) while Detoxify does not ($ρ= -0.23$, $p = .42$): toxicity scores fall as representational harm grows. We formalise harm laundering as a three-criteria test and provide a three-stage detection protocol applicable to any generative model. Within the OpenAI GPT lineage, toxicity score reduction is not a sufficient proxy for harm reduction.

Automatic speech recognition (ASR) systems exhibit unequal error rates across speaker groups, motivating interventions on their internal representations. We ask whether speaker-linked attributes that are linearly readable from pretrained ASR encoders yield useful directions for reducing group word-error-rate (WER) gaps. Across Whisper-medium, HuBERT-large, and Wav2Vec2-large on Common Voice and the Speech Accent Archive, we probe every encoder layer for metadata-derived sex/gender, age, and native/accent labels; construct centroid and probe-derived directions; inject them at selected layers; and compare downstream probe trajectories with matched WER changes. Sex labels are highly decodable (best macro-F1 0.924--0.941), native/accent labels are also above chance (0.544--0.696), and age is weaker (0.354--0.397). Of 22 post-selected reruns, nine have 95% paired-bootstrap intervals entirely below zero, yet every absolute source-group WER reduction is below 0.7 percentage points. Conversely, a local target-class probe rate can rise from 8.09% to 99.87% while WER worsens. Linear readability is therefore neither evidence of causal use nor a reliable mitigation method. Our results motivate evaluating speech-bias interventions jointly at representation, propagation, and task levels.

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.

Gender bias in large vision-language models (LVLMs) undermines their fairness and reliability, compromising output trustworthiness. Current mitigation methods rely on training-phase adjustments or post-hoc calibration, but face limitations in dynamic visual bias mitigation. These include inability to capture real-time visual-textual incongruence, dependence on predefined gender bias taxonomies, and degraded cross-modal alignment with emergent bias patterns. To address these challenges, we propose ViD, a causally-inspired framework that analyzes attention mechanisms across five distinct patterns, revealing confounding effects from strong language priors. ViD demonstrates that visual-to-language cross-attention effectively suppresses bias while preserving general reasoning capabilities and text generation quality. ViD incorporates dual mechanisms: backdoor adjustment counters strong language priors, while refined token selection in decoding layers optimizes processing. This enhances model robustness and inference efficiency. Our integrated approach significantly mitigates gender bias across multidimensional social attributes in LVLMs, improving visual grounding and output fairness. Cross-benchmark validation shows ViD reduces gender bias by 14.7% on single-attribute evaluations (FACET) and achieves significant improvements on image captioning tasks (MS COCO), with gender bias score improving from 0.6708 to 0.9978 for LLaVA. Crucially, these improvements require no additional training overhead, making ViD a scalable and practical solution for bias mitigation in LVLMs.

Despite steady progress in face recognition, current face recognition models still suffer from significant demographic biases. While approaches for bias mitigation have been proposed, existing methods often impose constraints on the training procedure and result in the degradation of recognition accuracy. To address this issue, we here introduce a method that reduces racial bias in pre-trained face recognition models without compromising their accuracy. To this end, we model face embeddings of each person by von Mises-Fisher (MF) distribution. We next observe the dependency between demographic attributes and the density of MF distributions, and propose DenseFace, a probabilistic face matching procedure that accounts for differences in MF distributions. Our extensive experiments demonstrate DenseFace to consistently reduce racial bias in strong face recognition models varying in network architectures, training datasets and loss functions. Notably, DenseFace preserves recognition accuracy and requires no retraining of the underlying face recognition model. Our work also investigates previously adopted bias measures and makes suggestions.

Detecting sexism on the internet is a fundamentally subjective task; our team, VANGUARD, addresses this challenge in the EXIST 2026 Task 2 by proposing a human-centered multimodal framework that analyses and incorporates the psychological and demographic characteristics of human annotators into the detection pipeline. We fuse five input modalities through a cross-attention architecture with Feature-wise Linear Modulation conditioning. Meme text is extracted and visually described with Gemma 4, then augmented by automatic translation between English and Spanish with NLLB-200. Text and image representations are produced by LoRAadapted XLM-RoBERTa and CLIP encoders and fused with sensor features encoded by a pretrained autoencoder. To model annotator subjectivity, we frame Subtask 2.1 as a label distribution learning problem, optimizing a Kullback-Leibler divergence loss over the full annotator label distribution. At inference time, predictions are produced by soft-voting between the deep multimodal network and a complementary SVM trained on stylometric and physiological features. Our best submission ranks 29th out of 114 on Subtask 2.2 (source intention) under soft evaluation, and the normalized ICM scores remain above the baseline on Subtasks 2.1 and 2.2, indicating that annotator-centered conditioning contributes a usable signal. We release our full pipeline and analysis to support reproducible human-centered modeling.

Diffusion-based Large Vision-Language Models (dLVLMs) have recently emerged as a compelling alternative to autoregressive (AR) LVLMs, offering advantages in parallel decoding, bidirectional context, and controllable generation. Despite rapid progress, their reliability properties remain largely uncharacterized. We present the first systematic reliability evaluation of hallucination and bias in dLVLMs, benchmarking six diffusion models against competitive AR baselines across four dimensions. Our key findings are: (1) dLVLMs reverse the yes-bias of AR models in binary visual queries; (2) they achieve competitive hallucination rates yet exhibit degraded linguistic quality; (3) they collapse to near-zero accuracy on underrepresented racial groups with opposite-polarity gender bias; and (4) they exhibit accuracy collapse in multiple-choice settings when the correct option is shorter than its distractors, associated with a length prior that emerges at the first denoising step. Tokens committed at late denoising steps with low confidence further correlate with hallucinated content, pointing to a mechanistic signal unique to diffusion generation. These patterns vary across model families, suggesting reliability is shaped by the generative paradigm together with training data.

The language a legislature uses to debate women's rights, even in favour of them, encodes systematic patterns of sexism that persist across two centuries. In this work, we analyse 6,531 speeches over 200 years of UK parliamentary debate (Hansard, 1803-2005) by using large language models to classify a speaker's perspective towards women's suffrage and political representation, as well as analyse sexist speech in parliament from the lens of the Ambivalent Sexism Inventory. We also release this parliamentary dataset, an organized and metadata-enriched version of the publicly available Hansard Corpus optimized for computational social science research, with 6.7 million speeches across 1.2 million debates, with 89% gender-matching for speeches by MPs from the House of Commons. We find that 54% of speeches opposing women's representation contain sexist content, compared to 21% of speeches that are for the cause, and that the two sides use fundamentally different types of sexism: anti-suffrage rhetoric combines hostile and benevolent framing, while pro-suffrage sexism is overwhelmingly benevolent. Female MPs support women's political rights at 93% compared to 70% for male MPs, a gap that closes only after enfranchisement. Our findings are evidence that benevolent and hostile sexism are used in different rhetorical contexts in a manner consistent with the theory of Ambivalent Sexism.

Machine translation (MT) systems often fail to correctly translate gender, especially when converting from a gender-neutral language like English to a gendered target language such as Romanian. This bias results in translations that default to masculine forms or reinforce gender stereotypes. We propose a hybrid pipeline to mitigate this issue by combining large language model (LLM)-based gender classification with neural machine translation (NMT). Our system uses a fine-tuned LLM to detect the intended gender of target words in English sentences and insert inline gender hint tags. These tagged sentences are then passed to a Transformer model fine-tuned to generate morphologically correct Romanian translations. To support this, we introduce three novel datasets for gender disambiguation and translation. Our approach improves gender accuracy on the WinoMT and WinoGender benchmarks by over 40 percentage points compared to a baseline MT system. This is the first method to explicitly address and evaluate gender bias in English-Romanian MT using both LLM inference and tag-aware translation.

AI tools are being increasingly adopted in policing in the UK and worldwide. Racial bias is a known and well-documented risk, yet representatives of affected communities are rarely included in decisions about AI adoption. We present results from a mixed-stakeholder deliberation workshop bringing together 30 community representatives, police officers, and academics to assess the risks of 13 AI use cases in policing, with an explicit focus on racial bias. We found that participants were broadly open to AI adoption, rejecting only three use cases outright, most notably recidivism risk assessment, where objections targeted the premise rather than the implementation. Our analysis reveals that foregrounding racial equity did not narrow the deliberation. Instead, discussions gravitated toward a fundamental set of questions: does this tool actually work, will it deliver genuine benefit, and will that benefit extend to everyone? This integrated reasoning, reminiscent of the curb-cut effect in inclusive design, highlights the benefit of incorporating the racial bias lens into the risk-benefit analysis of AI use cases from the outset.

Large Language Models (LLMs) are increasingly used for automated fact-checking, yet their susceptibility to gender bias in this context remains underexplored. This study presents the first systematic investigation of gender bias in LLM-based fake news detection using real-world data. We augment the LIAR benchmark with three gender variants of speaker job titles (Neutral, Male, Female) for each statement to test whether veracity judgments vary solely based on gender presentation. Six state-of-the-art LLMs are evaluated across multiple bias and fairness metrics. All models exhibit gender sensitivity: 9.79%-35.13% of statements receive inconsistent labels across the three variants, with Male-Female comparisons showing 6.5%-23.6% flip rates. Two primary bias manifestations are identified: instability (inconsistent judgments) and directionality (systematic favoritism). Five models show statistically significant directional effects, with the strongest effects displaying male-skeptic patterns. These findings demonstrate that gender bias undermines both reliability and fairness in LLM-based fake news detection, highlighting the need for bias-aware evaluation and mitigation strategies. The augmented dataset is publicly released to support future research.

When people label text for sexism, they often disagree, and not because some of them are wrong: they genuinely perceive sexism differently. Most NLP systems discard this disagreement by collapsing it into a majority vote. We propose the Multi-Agent Perspectivist Preference Optimization (MAP-PO) framework to keep these different perspectives. On the EXIST 2024 dataset of labeled English and Spanish tweets, we first cluster annotators by their labeling behavior rather than their demographic attributes. We then fine-tune one Large Language Model agent per cluster to reproduce that cluster's annotation behavior, and coordinate the agents with preference optimization that combines individual and team-level rewards. We evaluate MAP-PO in four settings defined by two languages and two backbone language models, asking whether each agent reproduces the annotations of its own cluster and whether the agents together reproduce the majority label. Two findings hold in all four settings. First, without fine-tuning the agents behave almost identically, so cluster-specific training is necessary. Second, we show that training each agent only on the labels of its own cluster pushes the agents far beyond the clusters they should represent, while adding a shared team-level training signal consistently keeps each agent calibrated to its cluster.

AI-based recruitment systems that rely on machine learning models trained on historical CV data, risk perpetuating and amplifying social biases. A key challenge arises in unstructured CV text, where pre-trained language model embeddings may infer sensitive attributes such as gender even after explicit indicators are removed. In this paper, we evaluate nine pre-trained embedding models on the synthetic FairCVdb dataset, analyzing the informativeness of their embeddings for applicant scoring and their susceptibility to gender leakage, on both original and gender-scrubbed biographies. We further use a multi-task adversarial learning framework with gradient reversal to predict applicant suitability while suppressing gender information from learned representations. Finally, we use a multi-objective Pareto-front-based model selection to balance predictive utility and fairness. Our experimental results show that explicit gender scrubbing substantially reduces but does not eliminate gender leakage, while adversarial learning improves fairness mainly on original biographies and acts as a complementary strategy rather than a substitute for text-level debiasing.

The proliferation of internet memes has introduced new complexities to automated content moderation, particularly in detecting misogyny. Memes often rely on a semantic clash between visual and textual modalities, where hateful intent is implicit and culturally grounded. This paper presents GeoMVC (Geometric Interaction and Multi-View Consensus), developed for the CC-MMD Grand Challenge at ICMI 2026. To address the limitations of static feature concatenation, a Geometric Interaction Layer is proposed that models cross-modal alignment via Hadamard products and cosine similarity between frozen visual and textual embeddings. We further mitigate distribution shifts caused by noisy OCR and code-mixed transliteration through a Multi-View Consensus strategy, aggregating predictions across raw, length-filtered, and English-translated text views. The system achieved Rank 2 in the Malayalam partition (Macro F1: 0.892) and Rank 3 in the Chinese partition (Macro F1: 0.895) on Task A, while securing Rank 5 in the Tamil partition (Macro F1: 0.521). A detailed error analysis on the development partition highlights open challenges in modeling localized transliteration and code-mixed sarcasm across Dravidian and Chinese cultural contexts.

Audio deepfake detection models determine whether speech is genuine or artificially generated, but high overall accuracy can mask substantial performance disparities across demographic groups. In this work, we investigate gender bias in audio deepfake detection using the ASVspoof5 dataset. We use ASVspoof5 under a controlled custom split designed to isolate gender-composition effects. We train attack-specific models on nine training sets with different gender compositions, ranging from female-only to male-only. We use a ResNet18 classifier with LogSpectrogram and WavLM-Base+ features, and we evaluated six post-hoc threshold calibration methods. Experimental results show that training data composition strongly predicts bias direction, with the underrepresented gender performing worse at test time. WavLM-Base+ features are shown to produce gender performance gaps 3.0 to 4.3 times larger than LogSpectrogram under identical training conditions, and balanced training is found to reduce LogSpectrogram bias but leave WavLM bias largely intact. Moreover, all six calibration strategies, including Oracle calibration with full test-set label access, leave the Equal Error Rate gap unchanged at 1.317 pp, confirming that threshold adjustment cannot correct underlying score distribution disparities. Overall, these findings suggest that gender fairness in audio deepfake detection must be addressed at training time, as post-hoc methods can only partially mitigate the resulting disparities

We audit fourteen mainstream large language models (LLMs) for hiring discrimination using the paired-resume methodology of Kline, Rose, and Walters (2022). The sole 2023-vintage model reproduces the pro-White callback gap documented in field experiments on labor market discrimination ($+2.12$ pp, significant at the 1% level). Every model released in 2024 or after shows either a null gap or a significant pro-Black reversal (up to $-3.01$ pp). The same pattern holds on the gender axis. Based on 24,024 paired postings per model across 14 models, our results document a reversal in the direction of algorithmic hiring bias across model generations.

Artificial intelligence (AI) can enhance what people who use augmentative and alternative communication (AAC) are able to do with their systems. However, evaluating AI-powered AAC interfaces can be difficult. People are intersectional beings and current evaluation metrics can struggle to capture the multifaceted and nuanced desires people may have for their AAC. We explore the complicated nature of six AAC problem spaces, explore how AI might be used in these spaces, and suggest more robust methods of evaluation that take the intersectional nuances of people into account. We also discuss broader issues that arise across these problem spaces and how they could be addressed using our proposed evaluation methods.

Large language models (LLMs) are increasingly deployed in hiring workflows, yet most research on gender bias in LLM hiring decisions has focused on English-language, Western-format resumes. This study examines whether pro-female gender bias extends to a Japanese corporate context and evaluates two practical mitigation strategies. Using a counterfactual resume design with 60 Japanese rirekisho-format resumes, 12 name pairs selected on linguistically grounded gender-signal criteria, and five state-of-the-art LLMs (Claude Sonnet 4.6, GPT-4o, DeepSeek-V3, Gemini 2.5 Flash, Llama 3.3 70B), we conducted 43,200 API calls across baseline, prompt instruction, and privacy filter conditions. A crossed random-effects linear mixed model confirms a significant pro-female bias across all five models, replicating Western findings in a non-Western context. A prompt-level gender-neutrality instruction produces no meaningful reduction in bias. A name-reliance analysis formally identifies the candidate name as the primary gender channel: removing the name from the prompt reduces the female effect by nearly its full magnitude. An unexpected incompatibility between the privacy filter and GPT-4o's content safety filter, resulting in a 42% refusal rate, highlights a practical deployment challenge for name anonymization in LLM-assisted recruitment pipelines.

Gender bias in LLMs has been studied extensively in model outputs, with biased prompts shown to amplify stereotyped generations. Whether such bias propagates into text produced by humans who use these systems, however, remains underexplored. We investigate whether gender bias in an LLM writing assistant transfers into career plan essays written by students. We first verify that a gender-biased prompt induces gender-differentiated language in LLM-generated essays, while a neutral prompt does not. We then recruited participants (N = 123) in a controlled environment to write career plan essays for paired biographical profiles differing only in gender under three conditions: no AI assistance, neutral LLM assistance, or gender-biased LLM assistance. Students in the biased condition produced essays with a significantly larger agentic gap and more gender-stereotypic occupation suggestions than those in the control and neutral conditions. Our results also reveal that this bias transfer is asymmetric: agency is suppressed in female-target essays while male-target writing remains largely unaffected. Our findings highlight the risk of bias propagation in AI-assisted writing, calling for fairness-aware design in educational AI tools.

Gender bias in AI-generated stories is a well-documented problem. While much attention has been paid to reducing or mitigating this bias, it is not always clear whether interventions produce genuinely fairer results. To investigate this issue, we examine how large language models (LLMs) handle gender assignment in a narrative context that is popular, highly ambiguous, and also known to closely reproduce human stereotypes: stories about talking animals. We prompt six leading LLMs to complete an English-language story about seven different anthropomorphic animal characters whose gender is unstated. We additionally iterate with four different narrative settings and a range of model temperatures. Across the 23.8K stories, we find that models frequently avoid gendering the animal character in the story (19% on average) or use gender-neutral language like "it" or "its" (38.2% on average). However, when gender is assigned, there is a significant masculine bias. Feminine animal characters are virtually absent, present in just 2.2% of stories vs. 40.6% that feature masculine characters. Our findings point to a broader argument: neutrality bites. In other words, models that prioritize neutrality to address social bias may actually contribute to the erasure of marginalized perspectives and identities. We suggest that alternative strategies beyond neutrality need to be pursued, such as ones that more equally distribute social possibilities across imagined subjects.

Debiasing methods based on principal component analysis (PCA) are broadly used to reduce gender bias in word embeddings used in LLMs, yet it remains unclear what aspects of bias they actually remove and how destructive this process is. These methods are based on the understanding that bias resides in a low-dimensional subspace, with the assumption that most of it can be captured by a few principal components. In this work, we conduct a systematic geometric analysis of PCA-based gender debiasing and investigate what is actually removed from the embedding space. Our experiments across multiple embeddings show that direct gender bias is primarily concentrated in the first principal component, supporting the low-rank bias hypothesis. However, associative bias measured by WEAT does not align with these principal directions and is instead spread across multiple embedding dimensions. Furthermore, as expected, we demonstrate that removing an increasing number of principal components leads to a consistent degradation of the embedding geometry, affecting semantic structure and vector relationships. These results reveal that PCA-based debiasing operates as a trade-off: while it effectively reduces certain forms of direct bias, it fails to eliminate distributed associations and introduces geometric distortion. Moreover, there is no universal optimal level of debiasing, as the balance between bias reduction and semantic preservation depends on the chosen metric and embedding. Overall, our findings suggest that bias in word embeddings is not purely low-rank and that simple subspace removal methods may be insufficient for comprehensive debiasing.

Large language models (LLMs) are rapidly assuming an intermediary role in housing search through the integration of listing platforms within conversational interfaces, mediating access to information, search, and recommendations within urban settings. We expand on prior work on racial steering in LLMs by conducting a behavioral audit of seven open-weight and closed-source LLMs across four U.S. cities, testing location recommendations across three iterative prompting conditions that progressively add lifestyle preference context and reflect fair housing paired-testing methodologies. We find that steering is an emergent behavior of the model's interpretive license rather than primarily a static property. Steering results from the interaction of a user's identity, preference articulation, and the spatial logic that a model has internalized about learned representations of place, preference, and opportunity in a given city, and how different types of users relate to it. While steering was present, it was not uniform in direction or magnitude across evaluated conditions. Preference-conditioned testing often increased or reconfigured the number of models that exhibited steering behaviors relative to baseline conditions, suggesting that LLMs may interpret what the same housing preference means differently depending on the racial identity of the user. Our findings also demonstrate that the city is not a neutral testing unit for LLM evaluation in place-based sectors, and results from one local market cannot be assumed to generalize to another. Local and domain expertise will be required in the housing sector to ensure that legal and institutional commitments to fair housing are not undermined while adopting AI tools that mediate spatial access.

We audit six large language models (LLMs) for gender stereotyping across English, Korean, Chinese, and Japanese. Three were developed primarily for English-language use (Claude, GPT, Gemini) and three for East Asian use (DeepSeek, Syn-Pro, HyperCLOVA X). We adopt the HEXACO-100 personality inventory and anchor each model against a cross-cultural human dataset spanning 48 countries to ask not whether LLMs are biased, but how far their gender attributions drift from the populations they are deployed among. Our findings show that their stereotyping spans a range roughly 2.5 times wider than the entire cross-country range found in humans, and the effect can compound across languages. One English-centric model, prompted in Korean, reached 5 times the local baseline, even when the prompt stated the candidate had already been hired, which often dampens human stereotyping. To characterize such behaviors without ranking them, we introduce a four-pattern framework -- concordance, suppression, reorganization, and amplification -- across 24 (model x language) cells. Item-level analysis reveals that translation does not just rescale stereotypes, but changes the attributes tied to it, hiding significant rearrangement under the surface while appearing well-calibrated. Our results ultimately suggest that no single debiasing pipeline is likely to address bias evenly across linguistic boundaries.

Language models (LMs) can produce gendered language and stereotypes even when given neutral prompts. Most prior work on gender bias in LMs primarily examines gender through a binary lens (feminine vs. masculine), with limited attention to gender-neutral forms, such as they/them pronouns or neutrally phrased job titles. How gender-related signals are encoded in the internal representations of LMs remains an open question. In this work, we study gender-specific neurons in LMs across three categories: feminine, masculine, and gender-neutral. We propose a neuron-level intervention method to identify neurons that are strongly tied to each gender category. We then test these neurons through controlled generation, showing that activating or masking gender-related neurons can steer a sentence toward a target gender form while preserving its original meaning. To evaluate the effectiveness of our gender-intervention approach, we curate two datasets with controlled sentences labeled across all three gender categories and validate the data quality through human evaluation. Experiments on two open-source LMs show that gender-specific neurons are not evenly distributed across model layers; instead, they concentrate heavily in the earliest layers with smaller contributions from later layers. Compared to existing methods, our method achieves more precise gender control, with less leakage into non-target gender categories and stable output quality through two evaluation criteria. Overall, our work examines how gender is encoded in LMs and provides a simple yet effective approach toward controlled gender intervention for both neuron intervention evaluation and gender bias mitigation. Code and datasets are available at: https://github.com/zhiwenyou103/Gender-Neuron-Intervention

While social media platforms, such as Twitter, provide a medium for large-scale opinion sharing during news events, it is manually impossible for individuals or media agencies to process the vast volume of content to identify key viewpoints. In order to resolve this, several automatic summarization techniques have been proposed to condense large collections of tweets into concise and informative summaries. However, these algorithms do not explicitly consider demographic fairness. Several existing research works have developed automated summarization approaches that can provide a holistic overview of the key aspects and major opinions shared on social media platforms related to a news event. However, these approaches do not explicitly consider different forms of demographic representation, such as gender, which can lead to biased summary representation. In this paper, we propose EquiSumm, which considers the gender aspect of the shared opinion to generate a summary, and our experimental analysis on two major datasets indicates the performance effectiveness with respect to existing research works.

Large language models (LLMs) are increasingly deployed in socially sensitive settings despite substantial documentation that they encode gender biases. Chain-of-Thought (CoT) prompting has been proposed as a bias-mitigation approach. However, existing evaluations primarily focus on changes in LLM benchmark performance, providing limited insight into whether apparent bias reductions reflect meaningful changes in a model's internal mechanisms. In this work, we investigate how CoT prompting affects gender bias in LLMs, combining benchmark-based evaluation with mechanistic interpretability techniques and reasoning chain failure analysis. Our results confirm a stereotypical bias present in LLM outputs across benchmarks, showing that CoT prompting does not consistently reduce the bias gap. Mechanistic analyses reveal that although CoT balances biased behavior in certain attention head clusters, gender bias remains embedded in hidden representations, indicating only superficial mitigation. Inspection of reasoning chains further suggests that these improvements stem from memorization and familiarity with the dataset rather than genuine understanding of bias.

Large-scale facial datasets like CelebA are widely used in computer vision, yet the cultural biases embedded in their labels remain underexplored. Fairness research has distinguished representational from allocational harms, but audits of computer vision datasets have mostly examined categorical labels, leaving open how such harms appear in learned features and model attention. This paper examines CelebA at three levels: dataset structure, learned feature weights, and spatial attention, focusing on how gendered double standards of ageing and beauty are encoded in the data and reproduced in model behaviour. First, hierarchical clustering of 202,599 images shows that the 39 attributes organise into latent trait bundles aligned with cultural archetypes: performative femininity (youth, makeup, adornment) and professional masculinity (ageing, facial hair, formal attire). Female faces, though more often rated attractive overall, incur steep penalties when assigned to ageing or masculine-coded clusters. Second, XGBoost with SHAP analysis reveal gender-specific effects, such as adiposity reducing attractiveness only for females. Third, Grad-CAM finds that predictions for female and younger male subgroups concentrate on mid-face cues, whereas predictions for older males drift toward peripheral cues such as hair and clothing. Older males attain the highest accuracy but the lowest average precision, indicating categorical exclusion of groups outside the dataset's evaluative templates. Cultural double standards thus pass from media representation into dataset labels, feature weights, and model attention, producing two representational harms: hyper-scrutiny of women under a narrow evaluative template, and exclusion of older men from the scheme entirely. Fairness metrics focused on performance disparities mask both, underscoring the need to address representational harm in fairness research.

Text-to-image (T2I) generative models are increasingly used to produce content for education, media, and public-facing communication, and are starting to be integrated into higher-impact pipelines. Since generated images tend to reinforce stereotypes, producing representational erasure via "default" depictions and shaping perceptions of who belongs in certain roles, a growing body of work has proposed metrics to quantify gender bias in T2I outputs. Yet existing evaluations remain fragmented. Metrics are often reported without a shared view of what they measure, what assumptions they entail, or how their results should be interpreted under different deployment contexts. This limits the usefulness of gender bias measurement for both technical auditing and emerging governance discussions. We propose a risk-aligned auditing framework for gender bias in T2I models composed of three constituents that connects risk categories, evaluation metrics, and harms. First, we identify risk-tiered use-case profiles aligned with the EU AI Act's risk categories to motivate why auditing expectations may vary with deployment contexts and stakeholder exposure. Second, we construct a metric catalog that consolidates gender-bias evaluation methods and organizes them in three measurement categories: gender prediction, embedding similarity, and downstream task. Third, we introduce a harm typology that maps context-dependent harm categories (e.g., representational, quality-of-service) to specific risk-tired scenarios. Finally, we introduce THUMB cards (Text-to-image Harms-informed Use-case-aligned Metrics of gender Bias) that help formulate auditing systematically by the incorporation of context, scenario and bias manifestation, harm hypotheses, and audit strategy.

Recent works have analyzed the impact of individual components of neural networks on gendered predictions, often with a focus on mitigating gender bias. However, mechanistic interpretations of gender tend to (i) focus on a very specific gender-related task, such as gendered pronoun prediction, or (ii) fail to distinguish between the production of factually gendered outputs (the correct assumption of gender given a word that carries gender as a semantic property) and gender biased outputs (based on a stereotype). To address these issues, we curate \gknow, a benchmark to assess gender knowledge and gender bias in language models across different types of gender-related predictions. \gknow allows us to identify and analyze circuits and individual neurons responsible for gendered predictions. We test the impact of neuron ablation on benchmarks for disentangling stereotypical and factual gender (DiFair and the test set of GKnow), as well as StereoSet. Results show that gender bias and factual gender are severely entangled on the level of both circuits and neurons, entailing that ablation is an unreliable debiasing method. Furthermore, we show that benchmarks for evaluating gender bias can hide the decrease in factual gender knowledge that accompanies neuron ablation. We curate GKnow as a contribution to the continuous development of robust gender bias benchmarks.

Human label variation has been established as a central phenomenon in NLP: the perspectives different annotators have on the same item need to be embraced. Data collection practices thus shifted towards increasing the annotator numbers and releasing disaggregated datasets, harmful language being most resourced due to its high subjectivity. While this resulted in rich information about \textit{who} annotated (sociodemographics, attitudes, etc.), the \textit{what} (e.g., linguistic properties of items), and their interplay has received little attention. We present the first large-scale analysis of four reference datasets for harmful language detection, bringing together annotator characteristics, linguistic properties of the items, and their interactions in a statistically informed picture. We find that interactions are crucial, revealing intersectional effects ignored in previous work, and that a strong role is played by lexical cues and annotator attitudes. Effect patterns, however, vary considerably across datasets. This urges caution about generalization and transferability.

Emergency department triage assigns patients an acuity score that determines treatment priority, and clinical evidence documents persistent gender disparities in human acuity assessment. As hospitals pilot large language models (LLMs) as triage decision support, a critical question is whether these models reproduce or mitigate known biases. We present EQUITRIAGE, a fairness audit of LLM-based ESI assignment evaluating five models (Gemini-3-Flash, Nemotron-3-Super, DeepSeek-V3.1, Mistral-Small-3.2, GPT-4.1-Nano) across 374,275 evaluations on 18,714 MIMIC-IV-ED vignettes under four prompt strategies. Of 9,368 originals, 9,346 are paired with a gender-swapped counterfactual. All five models produced flip rates above a pre-registered 5% threshold (9.9% to 43.8%). Two showed directional female undertriage (DeepSeek F/M 2.15:1, Gemini 1.34:1); two were near-parity; one had high sensitivity with weak male-direction asymmetry. DeepSeek's directional bias coexisted with a low outcome-linked calibration gap (0.013 against MIMIC-IV admission), a Chouldechova-style dissociation between within-group calibration and between-pair counterfactual invariance. Demographic blinding reduced Gemini's flip rate to 0.5%; an age-preserving blind variant left DeepSeek with residual F/M 1.25, implicating age as a residual channel. Chain-of-thought prompting degraded accuracy for all five models. A two-model ablation reveals opposite underlying mechanisms for the same directional phenotype: in Gemini the signal is emergent in the combined name+gender swap, while in DeepSeek the gender token alone carries it. EQUITRIAGE shows that group parity, counterfactual invariance, and gender calibration are distinct fairness properties, that intervention effectiveness is model-dependent, and that per-model counterfactual auditing should precede clinical deployment.

Counterfactual prompting (i.e., perturbing a single factor and measuring output change) is widely used to evaluate things like LLM bias and CoT faithfulness. But in this work we argue that observed effects cannot be attributed to the targeted factor without accounting for baseline ``meaning-preserving'' modifications to text that establish general model sensitivity. This is because every counterfactual edit is a compound treatment that bundles the variable of interest with incidental surface-form variation; this violates treatment variation irrelevance. We observe prediction flip rates on MedQA of 14.9% when we surgically change patient gender. However, this is statistically indistinguishable from the flip rates induced by simply paraphrasing inputs (14.1%). In this case, it would therefore be unwarranted to conclude that the LLM is especially sensitive to patient gender. To account for this and robustly measure the effects of targeted interventions, we propose a framework in which we compare (via statistical testing) differences observed under target interventions to those induced by paraphrasing inputs. We then use this framework to revisit a analysis done on the MedPerturb dataset, which reported evidence of model sensitivity to patient demographics and stylistic cues. We find that these effects largely dissipate when we account for general model sensitivity, with only 5 of 120 tests reaching statistical significance. Applying the same framework to occupational biography classification, we detect clearly significant directional gender bias, showing that the framework identifies real directional effects even when they are small. We evaluate a range of metrics -- aggregate, per-sample distributional, and regression -- and find that per-sample metrics are dramatically more powerful than aggregate metrics and regression powerfully and uniquely characterizes effect direction and magnitude.

Bias in medical AI is often framed as a problem of representation. However, in image-based tasks such as fetal ultrasound, performance disparities can arise even when representation is adequate, because predictive accuracy depends strongly on image quality. Image quality is shaped by acquisition conditions and operator expertise, as well as patient-dependent factors such as maternal body mass index (BMI), all of which may correlate with sensitive demographic features. Consequently, observed disparities may reflect the combined influence of demographic, clinical, and acquisition-related factors rather than data imbalance alone, and may obscure underlying interaction or confounding effects. We propose a structured framework to explore and detect intersectional bias, combining unsupervised slice discovery, systematic factor-wise analysis, and targeted intersectional evaluation. In a case study of over 94{,}000 ultrasound images for fetal weight estimation, we analyze bias in a state-of-the-art deep learning (DL) model and the clinical standard Hadlock, a regression formula using biometric measurements. Pixel spacing (PS) -- a parameter considered suboptimal in current acquisition protocols -- emerged as a consistent driver of performance differences, with higher PS associated with improvements of up to 24% in selected subgroups for both models. Because PS is often adapted in cases of high BMI or low gestational age (GA), this effect carries a substantial risk of confounding. Our intersectional analysis revealed that part of the PS-associated signal is explained by GA, while PS-related improvements persist across BMI strata, highlighting the importance of acquisition-aware and interaction-aware evaluation in medical AI fairness research.

Fairness in machine learning remains challenging due to its ethical complexity, the absence of a universal definition, and the need for context-specific bias metrics. Existing methods still struggle with intersectionality, multiclass settings, and limited flexibility and generality. To address these gaps, we introduce MIFair, a unified framework for bias assessment and mitigation based on mutual information. MIFair provides a flexible metric template and an in-processing mitigation method inspired by the Prejudice Remover, defining group fairness as statistical independence between prediction-derived variables and sensitive attributes. We further strengthen its information-theoretic foundation by establishing equivalences with widely used fairness notions such as independence and separation. MIFair naturally supports intersectionality, complex subgroup structures, and multiclass classification and employs regularization-based training to reduce bias according to the selected metric. Its key advantage is its versatility: it consolidates diverse fairness requirements into a single coherent framework, enabling consistent benchmarking and simplifying practical use. Experiments on real-world tabular and image datasets show that MIFair effectively reduces bias, including previously unaddressed multi-attribute scenarios, while maintaining strong predictive performance across the evaluated settings.

Text-to-Image (T2I) models have shown promising utility across various domains. However, such models are also amplifying harmful societal biases in their outputs. In the context of South Asia, recent work has shown caste biases and stereotypes are being perpetuated through Generative AI (GenAI) systems. While this research offers extremely relevant insight into invisibilized narratives of caste discrimination through the GenAI system, they often treat caste as an identity category. Therefore, in this work we shift our ontology to focus on the relational aspect of caste. This enables us to develop a more nuanced understanding of the mechanics of caste discrimination by and through T2I models. Combining an algorithmic audit with critical discourse analysis, we draw on a conceptual frame challenging Brahminical Normativity to show how caste biases are perpetuated beyond the simple binaries of upper vs lower-caste categories. Our contributions are two-fold. Beyond challenging the categorical understanding of caste as a category, we propose an anti-caste approach to tackle the issue of caste bias and fairness in AI systems.

AI risk assessment is the primary tool for identifying harms caused by AI systems. These include intersectional harms, which arise from the interaction between identity categories (e.g., class and skin tone) and which do not occur, or occur differently, when those categories are considered separately. Yet existing AI risk assessments are still built around isolated identity categories, and when intersections are considered, they focus almost exclusively on race and gender. Drawing on a large-scale analysis of documented AI incidents, we show that AI harms do not occur one identity category at a time. Using a structured rubric applied with a Large Language Model (LLM), we analyze 5,300 reports from 1,200 documented incidents in the AI Incident Database, the most curated source of incident data. From these reports, we identify 1,513 harmed subjects and their associated identity categories, achieving 98% accuracy. At the level of individual categories, we find that age and political identity appear in documented AI harms at rates comparable to race and gender. At the level of intersecting categories, harm is amplified up to three times at specific intersections: adolescent girls, lower-class people of color, and upper-class political elites. We argue that intersectionality should be a core component of AI risk assessment to more accurately capture how harms are produced and distributed across social groups.

Large Language Models (LLMs) are increasingly deployed in persona-driven applications such as education, customer service, and social platforms, where models are prompted to adopt specific personas when interacting with users. While persona conditioning can improve user experience and engagement, it also raises concerns about how personality cues may interact with gender biases and stereotypes. In this work, we present a controlled study of persona-conditioned story generation in English and Hindi, where each story portrays a working professional in India producing context-specific artifacts (e.g., lesson plans, reports, letters) under systematically varied persona gender, occupational role, and personality traits from the HEXACO and Dark Triad frameworks. Across 23,400 generated stories from six state-of-the-art LLMs, we find that personality traits are significantly associated with both the magnitude and direction of gender bias. In particular, Dark Triad personality traits are consistently associated with higher gender-stereotypical representations compared to socially desirable HEXACO traits, though these associations vary across models and languages. Our findings demonstrate that gender bias in LLMs is not static but context-dependent. This suggests that persona-conditioned systems used in real-world applications may introduce uneven representational harms, reinforcing gender stereotypes in generated educational, professional, or social content.

Quality Estimation (QE) aims to assess machine translation quality without reference translations, but recent studies have shown that existing QE models exhibit systematic gender bias. In particular, they tend to favor masculine realizations in gender-ambiguous contexts and may assign higher scores to gender-misaligned translations even when gender is explicitly specified. To address these issues, we propose FairQE, a multi-agent-based, fairness-aware QE framework that mitigates gender bias in both gender-ambiguous and gender-explicit scenarios. FairQE detects gender cues, generates gender-flipped translation variants, and combines conventional QE scores with LLM-based bias-mitigating reasoning through a dynamic bias-aware aggregation mechanism. This design preserves the strengths of existing QE models while calibrating their gender-related biases in a plug-and-play manner. Extensive experiments across multiple gender bias evaluation settings demonstrate that FairQE consistently improves gender fairness over strong QE baselines. Moreover, under MQM-based meta-evaluation following the WMT 2023 Metrics Shared Task, FairQE achieves competitive or improved general QE performance. These results show that gender bias in QE can be effectively mitigated without sacrificing evaluation accuracy, enabling fairer and more reliable translation evaluation.

Large language models (LLMs) are increasingly used in clinical settings, raising concerns about racial bias in both generated medical text and clinical reasoning. Existing studies have identified bias in medical LLMs, but many focus on single models and give less attention to mitigation. This study uses the EU AI Act as a governance lens to evaluate five widely used LLMs across two tasks, namely synthetic patient-case generation and differential diagnosis ranking. Using race-stratified epidemiological distributions in the United States and expert differential diagnosis lists as benchmarks, we apply structured prompt templates and a two-part evaluation design to examine implicit and explicit racial bias. All models deviated from observed racial distributions in the synthetic case generation task, with GPT-4.1 showing the smallest overall deviation. In the differential diagnosis task, DeepSeek V3 produced the strongest overall results across the reported metrics. When embedded in an agentic workflow, DeepSeek V3 showed an improvement of 0.0348 in mean p-value, 0.1166 in median p-value, and 0.0949 in mean difference relative to the standalone model, although improvement was not uniform across every metric. These findings support multi-metric bias evaluation for AI systems used in medical settings and suggest that retrieval-based agentic workflows may reduce some forms of explicit bias in benchmarked diagnostic tasks. Detailed prompt templates, experimental datasets, and code pipelines are available on our GitHub.

Large language models (LLMs) are being increasingly used in urban planning, but since gendered space theory highlights how gender hierarchies are embedded in spatial organization, there is concern that LLMs may reproduce or amplify such biases. We introduce SPAGBias - the first systematic framework to evaluate spatial gender bias in LLMs. It combines a taxonomy of 62 urban micro-spaces, a prompt library, and three diagnostic layers: explicit (forced-choice resampling), probabilistic (token-level asymmetry), and constructional (semantic and narrative role analysis). Testing six representative models, we identify structured gender-space associations that go beyond the public-private divide, forming nuanced micro-level mappings. Story generation reveals how emotion, wording, and social roles jointly shape "spatial gender narratives". We also examine how prompt design, temperature, and model scale influence bias expression. Tracing experiments indicate that these patterns are embedded and reinforced across the model pipeline (pre-training, instruction tuning, and reward modeling), with model associations found to substantially exceed real-world distributions. Downstream experiments further reveal that such biases produce concrete failures in both normative and descriptive application settings. This work connects sociological theory with computational analysis, extending bias research into the spatial domain and uncovering how LLMs encode social gender cognition through language.

Generative AI models reproduce the human biases in their training data and further amplify them through mechanisms such as mode collapse. The loss of diversity produces homogenization, which not only harms the minoritized but impoverishes everyone. We argue homogenization should be a central concern in AI safety. To meaningfully characterize homogenization in Large Language Models (LLMs), we introduce a framework that allows stakeholders to encode their context and value system. We illustrate our approach with an experiment that surfaces gender bias in an LLM (Claude 3.5 Haiku) on an open-ended story prompt. Building from queer theory, we formalize homogenization in terms of normativity. Borrowing language from feminist theory, we introduce the concept of xeno-reproduction as a class of tasks for mitigating homogenization by promoting diversity. Our work opens a collaborative line of research that seeks to understand and advance diversity in AI.

Instruct-based large language models (LLMs) have been shown to propagate and even amplify gender bias when prompted with contextually constrained instructions (e.g., writing a text from a description or selecting a gendered pronoun). However, little attention has been paid to biases in responses to contextually unconstrained (generic) instructions conveyed by gendered language, particularly masculine generics (MG). MG, found in many gender-marked languages, denote the use of the masculine gender as a supposedly neutral reference to mixed-gender groups or individuals whose gender is unknown or non-binary. Yet, psycholinguistic studies demonstrate that MG are not neutral and systematically induce gender bias. This study investigates how both local and proprietary LLMs are MG-biased when responding to generic prompts in French, examining LLMs' MG bias rates and use of gender-fair language (GFL). We create a 16k+ human noun database from existing lexical resources and evaluate six LLMs on four instruction-response datasets under two conditions: prompts with and without MG. Overall, we find that $\approx$27.57% of LLMs' responses to MG-filtered generic instructions are MG-biased ($\approx$78.55% with MG-containing prompts). Moreover, we find that LLMs rarely use GFL spontaneously. These findings highlight the persistence of MG bias in LLM outputs and models' limited tendency towards GFL strategies.

During training, Large Language Models (LLMs) learn social regularities that can lead to gender bias in downstream applications. Most mitigation efforts focus on reducing bias in generated outputs, typically evaluated on structured benchmarks, which raises two concerns: output-level evaluation does not reveal whether alignment modifies the model's underlying representations, and structured benchmarks may not reflect realistic usage scenarios. We propose a unified framework to jointly analyze intrinsic and extrinsic gender bias in LLMs using identical neutral prompts, enabling direct comparison between gender-related information encoded in internal representations and bias expressed in generated outputs. Contrary to prior work reporting weak or inconsistent correlations, we find a consistent association between latent gender information and expressed bias when measured under the unified protocol. We further examine the effect of alignment through supervised fine-tuning aimed at reducing gender bias. Our results suggest that while the latter indeed reduces expressed bias, measurable gender-related associations are still present in internal representations, and can be reactivated under adversarial prompting. Finally, we consider two realistic settings and show that debiasing effects observed on structured benchmarks do not necessarily generalize, e.g., to the case of story generation.