We introduce VisoGender, a novel dataset for benchmarking gender bias in vision-language models. We focus on occupation-related biases within a hegemonic system of binary gender, inspired by Winograd and Winogender schemas, where each image is associated with a caption containing a pronoun relationship of subjects and objects in the scene. VisoGender is balanced by gender representation in professional roles, supporting bias evaluation in two ways: i) resolution bias, where we evaluate the difference between pronoun resolution accuracies for image subjects with gender presentations perceived as masculine versus feminine by human annotators and ii) retrieval bias, where we compare ratios of professionals perceived to have masculine and feminine gender presentations retrieved for a gender-neutral search query. We benchmark several state-of-the-art vision-language models and find that they demonstrate bias in resolving binary gender in complex scenes. While the direction and magnitude of gender bias depends on the task and the model being evaluated, captioning models are generally less biased than Vision-Language Encoders.

Many interpretation methods for neural models in natural language processing investigate how information is encoded inside hidden representations. However, these methods can only measure whether the information exists, not whether it is actually used by the model. We propose a methodology grounded in the theory of causal mediation analysis for interpreting which parts of a model are causally implicated in its behavior. The approach enables us to analyze the mechanisms that facilitate the flow of information from input to output through various model components, known as mediators. As a case study, we apply this methodology to analyzing gender bias in pre-trained Transformer language models. We study the role of individual neurons and attention heads in mediating gender bias across three datasets designed to gauge a model's sensitivity to gender bias. Our mediation analysis reveals that gender bias effects are concentrated in specific components of the model that may exhibit highly specialized behavior.

This paper examines how international AI governance frameworks address gender issues and gender-based harms. The analysis covers binding regulations, such as the EU AI Act; soft law instruments, like the UNESCO Recommendations on AI Ethics; and global initiatives, such as the Global Partnership on AI (GPAI). These instruments reveal emerging trends, including the integration of gender concerns into broader human rights frameworks, a shift toward explicit gender-related provisions, and a growing emphasis on inclusivity and diversity. Yet, some critical gaps persist, including inconsistent treatment of gender across governance documents, limited engagement with intersectionality, and a lack of robust enforcement mechanisms. However, this paper argues that effective AI governance must be intersectional, enforceable, and inclusive. This is key to moving beyond tokenism toward meaningful equity and preventing reinforcement of existing inequalities. The study contributes to ethical AI debates by highlighting the importance of gender-sensitive governance in building a just technological future.

Interpretability can be implemented as a means to understand decisions taken by (black box) models, such as machine translation (MT) or large language models (LLMs). Yet, research in this area has been limited in relation to a manifested problem in these models: gender bias. With this research, we aim to move away from simply measuring bias to exploring its origins. Working with gender-ambiguous natural source data, this study examines which context, in the form of input tokens in the source sentence, influences (or triggers) the translation model choice of a certain gender inflection in the target language. To analyse this, we use contrastive explanations and compute saliency attribution. We first address the challenge of a lacking scoring threshold and specifically examine different attribution levels of source words on the model gender decisions in the translation. We compare salient source words with human perceptions of gender and demonstrate a noticeable overlap between human perceptions and model attribution. Additionally, we provide a linguistic analysis of salient words. Our work showcases the relevance of understanding model translation decisions in terms of gender, how this compares to human decisions and that this information should be leveraged to mitigate gender bias.

Large Vision-Language Models (LVLMs) have transformed image captioning, shifting from concise captions to detailed descriptions. We introduce LOTUS, a leaderboard for evaluating detailed captions, addressing three main gaps in existing evaluations: lack of standardized criteria, bias-aware assessments, and user preference considerations. LOTUS comprehensively evaluates various aspects, including caption quality (e.g., alignment, descriptiveness), risks (\eg, hallucination), and societal biases (e.g., gender bias) while enabling preference-oriented evaluations by tailoring criteria to diverse user preferences. Our analysis of recent LVLMs reveals no single model excels across all criteria, while correlations emerge between caption detail and bias risks. Preference-oriented evaluations demonstrate that optimal model selection depends on user priorities.

Audio-based depression detection models have demonstrated promising performance but often suffer from gender bias due to imbalanced training data. Epidemiological statistics show a higher prevalence of depression in females, leading models to learn spurious correlations between gender and depression. Consequently, models tend to over-diagnose female patients while underperforming on male patients, raising significant fairness concerns. To address this, we propose a novel Counterfactual Debiasing Framework grounded in causal inference. We construct a causal graph to model the decision-making process and identify gender bias as the direct causal effect of gender on the prediction. During inference, we employ counterfactual inference to estimate and subtract this direct effect, ensuring the model relies primarily on authentic acoustic pathological features. Extensive experiments on the DAIC-WOZ dataset using two advanced acoustic backbones demonstrate that our framework not only significantly reduces gender bias but also improves overall detection performance compared to existing debiasing strategies.

Unlike text, speech conveys information about the speaker, such as gender, through acoustic cues like pitch. This gives rise to modality-specific bias concerns. For example, in speech translation (ST), when translating from languages with notional gender, such as English, into languages where gender-ambiguous terms referring to the speaker are assigned grammatical gender, the speaker's vocal characteristics may play a role in gender assignment. This risks misgendering speakers, whether through masculine defaults or vocal-based assumptions. Yet, how ST models make these decisions remains poorly understood. We investigate the mechanisms ST models use to assign gender to speaker-referring terms across three language pairs (en-es/fr/it), examining how training data patterns, internal language model (ILM) biases, and acoustic information interact. We find that models do not simply replicate term-specific gender associations from training data, but learn broader patterns of masculine prevalence. While the ILM exhibits strong masculine bias, models can override these preferences based on acoustic input. Using contrastive feature attribution on spectrograms, we reveal that the model with higher gender accuracy relies on a previously unknown mechanism: using first-person pronouns to link gendered terms back to the speaker, accessing gender information distributed across the frequency spectrum rather than concentrated in pitch.

Text-to-image (T2I) diffusion models often exhibit gender bias, particularly by generating stereotypical associations between professions and gendered subjects. This paper presents SAE Debias, a lightweight and model-agnostic framework for mitigating such bias in T2I generation. Unlike prior approaches that rely on CLIP-based filtering or prompt engineering, which often require model-specific adjustments and offer limited control, SAE Debias operates directly within the feature space without retraining or architectural modifications. By leveraging a k-sparse autoencoder pre-trained on a gender bias dataset, the method identifies gender-relevant directions within the sparse latent space, capturing professional stereotypes. Specifically, a biased direction per profession is constructed from sparse latents and suppressed during inference to steer generations toward more gender-balanced outputs. Trained only once, the sparse autoencoder provides a reusable debiasing direction, offering effective control and interpretable insight into biased subspaces. Extensive evaluations across multiple T2I models, including Stable Diffusion 1.4, 1.5, 2.1, and SDXL, demonstrate that SAE Debias substantially reduces gender bias while preserving generation quality. To the best of our knowledge, this is the first work to apply sparse autoencoders for identifying and intervening in gender bias within T2I models. These findings contribute toward building socially responsible generative AI, providing an interpretable and model-agnostic tool to support fairness in text-to-image generation.

The blind application of machine learning runs the risk of amplifying biases present in data. Such a danger is facing us with word embedding, a popular framework to represent text data as vectors which has been used in many machine learning and natural language processing tasks. We show that even word embeddings trained on Google News articles exhibit female/male gender stereotypes to a disturbing extent. This raises concerns because their widespread use, as we describe, often tends to amplify these biases. Geometrically, gender bias is first shown to be captured by a direction in the word embedding. Second, gender neutral words are shown to be linearly separable from gender definition words in the word embedding. Using these properties, we provide a methodology for modifying an embedding to remove gender stereotypes, such as the association between the words receptionist and female, while maintaining desired associations such as between the words queen and female. Using crowd-worker evaluation as well as standard benchmarks, we empirically demonstrate that our algorithms significantly reduce gender bias in embeddings while preserving the its useful properties such as the ability to cluster related concepts and to solve analogy tasks. The resulting embeddings can be used in applications without amplifying gender bias.