Language models are increasingly mediating information access to end users, urging a systematic evaluation of their responses for a fair and reliable information ecosystem. Existing evaluations, however, are often topic-specific or synthetic, limiting their ability to capture the complexity of "in the wild" information-seeking queries and the risks present in model responses. To address this gap, we introduce WildSEEK, a manually annotated dataset of 3k information-seeking queries from real user interactions, and an evaluation framework for LLM-generated responses. WildSEEK includes annotations for risk-sensitive domains (e.g. health and financial information), and distinguishes factoid queries from analytical queries which seek responses beyond facts. We train classifiers on WildSEEK to analyze more than 1.8M realistic user queries. We find that over a third of information-seeking queries are high-risk and more often analytical. Our findings show that LLM responses fail more often in four criteria: sycophantic behavior, overreliance, a default US-centric perspective, and poor handling of vulnerable populations – with failure rates being mostly higher for analytical queries. By providing methods to monitor the reliability, safety, and fairness of LLM behavior, our dataset and evaluation framework offer an empirical foundation for the broader question of how these systems should behave as they take on a growing role in information access.

We investigate the extent to which research similarity between senior and junior researchers influences promotion in academia and study its implications in terms of gender diversity among faculty. Using data on the universe of job applications for tenure-track assistant professor positions in economics in Italy and exploiting NLP techniques (i.e., document embeddings) on the abstract of each publication of the scholars in our dataset, we propose a novel measure of research similarity, which can capture closeness in research topics, methodologies or policy relevance between candidates and members of selection committees. We show that the level of similarity is strongly associated with the winning probability. Moreover, while there are no gender differences in average similarity, maximum similarity with members of the selection committee is lower for female candidates. This gender gap disappears when similarity is calculated only focusing on female members of the committee. The results suggest that similarity bias in male-dominated environments can have implications for gender and research diversity.

Given Farsi's speaker base of over 127 million people and the growing availability of digital text, including more than 1.3 million articles on Wikipedia, it is considered a middle-resource language. However, this label quickly crumbles when the situation is examined more closely. We focus on three subjective tasks (Sentiment Analysis, Emotion Analysis, and Toxicity Detection) and find significant challenges in data availability and quality, despite the overall increase in data availability. We review 110 publications on subjective tasks in Farsi and observe a lack of publicly available datasets. Furthermore, existing datasets often lack essential demographic factors, such as age and gender, that are crucial for accurately modeling subjectivity in language. When evaluating prediction models using the few available datasets, the results are highly unstable across both datasets and models. Our findings indicate that the volume of data is insufficient to significantly improve a language's prospects in NLP.

People increasingly rely on Large Language Models (LLMs) for moral advice, which may influence humans' decisions. Yet, little is known about how closely LLMs align with human moral judgments. To address this, we introduce the Moral Dilemma Dataset, a benchmark of 1,618 real-world moral dilemmas paired with a distribution of human moral judgments consisting of a binary evaluation and a free-text rationale. We treat this problem as a pluralistic distributional alignment task, comparing the distributions of LLM and human judgments across dilemmas. We find that models reproduce human judgments only under high consensus; alignment deteriorates sharply when human disagreement increases. In parallel, using a 60-value taxonomy built from 3,783 value expressions extracted from rationales, we show that LLMs rely on a narrower set of moral values than humans. These findings reveal a pluralistic moral gap: a mismatch in both the distribution and diversity of values expressed. To close this gap, we introduce Dynamic Moral Profiling (DMP), a Dirichlet-based sampling method that conditions model outputs on human-derived value profiles. DMP improves alignment by 64.3

Recent advances in large language models (LLMs) demonstrate their potential as educational tutors. However, different tutoring strategies benefit different student personalities, and mismatches can be counterproductive to student outcomes. Despite this, current LLM tutoring systems do not take into account student personality traits. To address this problem, we first construct a taxonomy that links pedagogical methods to personality profiles, based on pedagogical literature. We simulate student-teacher conversations and use our framework to let the LLM tutor adjust its strategy to the simulated student personality. We evaluate the scenario with human teachers and find that they consistently prefer our approach over two baselines. Our method also increases the use of less common, high-impact strategies such as role-playing, which human and LLM annotators prefer significantly. Our findings pave the way for developing more personalized and effective LLM use in educational applications.

Personalized content moderation can protect users from harm while facilitating free expression by tailoring moderation decisions to individual preferences rather than enforcing universal rules. However, content moderation that is fully personalized to individual preferences, no matter what these preferences are, may lead to even the most hazardous types of content being propagated on social media. In this paper, we explore this risk using hate speech as a case study. Certain types of hate speech are illegal in many countries. We show that, while fully personalized hate speech detection models increase overall user welfare (as measured by user-level classification performance), they also make predictions that violate such legal hate speech boundaries, especially when tailored to users who tolerate highly hateful content. To address this problem, we enforce legal boundaries in personalized hate speech detection by overriding predictions from personalized models with those from a boundary classifier. This approach significantly reduces legal violations while minimally affecting overall user welfare. Our findings highlight both the promise and the risks of personalized moderation, and offer a practical solution to balance user preferences with legal and ethical obligations.

The rise of Artificial Intelligence (AI) language technologies, particularly generative AI (GenAI) chatbots accessible via conversational interfaces, is transforming digital interactions. While these tools hold societal promise, they also risk widening digital divides due to uneven adoption and low awareness of their limitations. This study presents the first comprehensive empirical mapping of GenAI adoption, usage patterns, and literacy in Italy, based on newly collected survey data from 1,906 Italian-speaking adults. Our findings reveal widespread adoption for both work and personal use, including sensitive tasks like emotional support and medical advice. Crucially, GenAI is supplanting other technologies to become a primary information source: this trend persists despite low user digital literacy, posing a risk as users struggle to recognize errors or misinformation. Moreover, we identify a significant gender divide – particularly pronounced in older generations – where women are half as likely to adopt GenAI and use it less frequently than men. While we find literacy to be a key predictor of adoption, it only partially explains this disparity, suggesting that other barriers are at play. Overall, our data provide granular insights into the multipurpose usage of GenAI, highlighting the dual need for targeted educational initiatives and further investigation into the underlying barriers to equitable participation that competence alone cannot explain.

Persistent societal biases like misogyny express themselves more often implicitly than through openly hostile language.However, previous misogyny studies have focused primarily on explicit language, overlooking these more subtle forms. We bridge this gap by examining implicit misogynistic expressions in English and Italian. First, we develop a taxonomy of social dynamics, i.e., the underlying communicative intent behind misogynistic statements in social media data. Then, we test the ability of nine LLMs to identify the social dynamics as a multi-label classification and text span selection: first LLMs must choose social dynamics given a prefixed list, then they have to explicitly identify the text spans that triggered their decisions. We also investigate the extent of using different learning settings: zero and few-shot, and prescriptive. Our analysis suggests that LLMs struggle to follow instructions and reason in all settings, mostly relying on semantic associations, recasting claims of emergent abilities.

Stories play a pivotal role in human communication, shaping beliefs and morals, particularly in children. As parents increasingly rely on large language models (LLMs) to craft bedtime stories, the presence of cultural and gender stereotypes in these narratives raises significant concerns. To address this issue, we present Biased Tales, a comprehensive dataset designed to analyze how biases influence protagonists’ attributes and story elements in LLM-generated stories. Our analysis uncovers striking disparities. When the protagonist is described as a girl (as compared to a boy), appearance-related attributes increase by 55.26%. Stories featuring non-Western children disproportionately emphasize cultural heritage, tradition, and family themes far more than those for Western children. Our findings highlight the role of sociocultural bias in making creative AI use more equitable and diverse.

Large language models (LLMs) are increasingly used in tasks that shape public discourse, yet concerns remain about their potential to reproduce harmful social biases. In this paper, we investigate how LLMs represent feminists in Italian, focusing on both implicit associations and explicit characterizations. We develop a controlled prompt-based evaluation framework that compares model responses to prompts about feminists with those about comparable groups (e.g., women, male/female activists). Using a combination of single-word autocompletion and descriptive prompts, we analyze the sentiment, stereotypes, and lexical patterns present in the generated outputs. Our findings reveal that prompts invoking public perception elicit markedly more negative and stereotypical language, with feminists been often described as aggressive or extremist. These traits are less attributed to 'women' or 'activists'. We also assess lexical hallucinations, noting a tendency towards generating stigmatizing neologisms. Last, we extract representative seed words from a corpus about feminism-related tweets and compute their semantic similarity to feminist(s) via contextualized word embeddings to uncover the models' implicit biases encoded in their internal semantic representations. The results show that the plural form 'femministe' is more tightly linked to politicized and negative framings.

The first international workshop on protecting women online, Towards a Safer Web for Women (TSWW), was held in conjunction with the Web Conference 2025. The workshop addresses the urgent and growing issue of online violence against women and girls, identifying and addressing the current challenges. As online and digital environments become increasingly hostile due to the proliferation of hate speech, deepfake abuse, and algorithmic gender bias, there is an urgent need for international and interdisciplinary collaboration to develop effective solutions and mitigation strategies. This workshop brings together researchers primarily from the computer science domain (e.g. natural language processing and machine learning experts) alongside experts in other domains (e.g. legislation and human behaviour experts), recognising the critical role of interdisciplinary collaboration in addressing this complex issue and encouraging knowledge exchange and innovation across fields. In this paper, we present an overview of the inaugural edition of the TSWW workshop.

Large language models (LLMs) are known to generate politically biased text, yet how such biases arise remains unclear. A crucial step toward answering this question is the analysis of training data, whose political content remains largely underexplored in current LLM research. To address this gap, we present in this paper an analysis of the pre- and post-training corpora of OLMO2, the largest fully open-source model released together with its complete dataset. From these corpora, we draw large random samples, automatically annotate documents for political orientation, and analyze their source domains and content. We then assess how political content in the training data correlates with models' stance on specific policy issues. Our analysis shows that left-leaning documents predominate across datasets, with pre-training corpora containing significantly more politically engaged content than post-training data. We also find that left- and right-leaning documents frame similar topics through distinct values and sources of legitimacy. Finally, the predominant stance in the training data strongly correlates with models' political biases when evaluated on policy issues. These findings underscore the need to integrate political content analysis into future data curation pipelines as well as in-depth documentation of filtering strategies for transparency.

As Natural Language Processing models become increasingly embedded in everyday life, ensuring that these systems can measure and mitigate bias is critical. While substantial work has been done to identify and mitigate gender bias in English, Farsi remains largely underexplored. This paper presents the first comprehensive study of gender bias in language models in Farsi across three tasks: emotion analysis, question answering, and hurtful sentence completion. We assess a range of language models across all the tasks in zero-shot settings. By adapting established evaluation frameworks for Farsi, we uncover patterns of gender bias that differ from those observed in English, highlighting the urgent need for culturally and linguistically inclusive approaches to bias mitigation in NLP.

In today’s digital age, hate speech and offensive speech online pose a significant challenge to maintaining respectful and inclusive online environments. This tutorial aims to provide attendees with a comprehensive understanding of the field by delving into essential dimensions such as multilingualism, counter-narrative generation, a hands-on session with one of the most popular APIs for detecting hate speech, fairness, and ethics in AI, and the use of recent advanced approaches. In addition, the tutorial aims to foster collaboration and inspire participants to create safer online spaces by detecting and mitigating hate speech.

This paper introduces the concept of actionability in the context of bias measures in natural language processing (NLP). We define actionability as the degree to which a measure’s results enable informed action and propose a set of desiderata for assessing it. Building on existing frameworks such as measurement modeling, we argue that actionability is a crucial aspect of bias measures that has been largely overlooked in the literature.We conduct a comprehensive review of 146 papers proposing bias measures in NLP, examining whether and how they provide the information required for actionable results. Our findings reveal that many key elements of actionability, including a measure’s intended use and reliability assessment, are often unclear or entirely absent.This study highlights a significant gap in the current approach to developing and reporting bias measures in NLP. We argue that this lack of clarity may impede the effective implementation and utilization of these measures. To address this issue, we offer recommendations for more comprehensive and actionable metric development and reporting practices in NLP bias research.

Language Models (LMs) have been shown to inherit undesired biases that might hurt minorities and underrepresented groups if such systems were integrated into real-world applications without careful fairness auditing. This paper proposes FairBelief, an analytical approach to capture and assess beliefs, i.e., propositions that an LM may embed with different degrees of confidence and that covertly influence its predictions. With FairBelief, we leverage prompting to study the behavior of several state-of-the-art LMs across different previously neglected axes, such as model scale and likelihood, assessing predictions on a fairness dataset specifically designed to quantify LMs' outputs' hurtfulness. Finally, we conclude with an in-depth qualitative assessment of the beliefs emitted by the models. We apply FairBelief to English LMs, revealing that, although these architectures enable high performances on diverse natural language processing tasks, they show hurtful beliefs about specific genders. Interestingly, training procedure and dataset, model scale, and architecture induce beliefs of different degrees of hurtfulness.

Machine learning (ML)-based content moderation tools are essential to keep online spaces free from hateful communication. Yet ML tools can only be as capable as the quality of the data they are trained on allows them. While there is increasing evidence that they underperform in detecting hateful communications directed towards specific identities and may discriminate against them, we know surprisingly little about the provenance of such bias. To fill this gap, we present a systematic review of the datasets for the automated detection of hateful communication introduced over the past decade, and unpack the quality of the datasets in terms of the identities that they embody: those of the targets of hateful communication that the data curators focused on, as well as those unintentionally included in the datasets. We find, overall, a skewed representation of selected target identities and mismatches between the targets that research conceptualizes and ultimately includes in datasets. Yet, by contextualizing these findings in the language and location of origin of the datasets, we highlight a positive trend towards the broadening and diversification of this research space.