Ensuring the safety of Generative AI requires a nuanced understanding of pluralistic viewpoints. In this paper, we introduce a novel data-driven approach for analyzing ordinal safety ratings in pluralistic settings. Specifically, we address the challenge of interpreting nuanced differences in safety feedback from a diverse population expressed via ordinal scales (e.g., a Likert scale). We define non-parametric responsiveness metrics that quantify how raters convey broader distinctions and granular variations in the severity of safety violations. Leveraging publicly available datasets of pluralistic safety feedback as our case studies, we investigate how raters from different demographic groups use an ordinal scale to express their perceptions of the severity of violations. We apply our metrics across violation types, demonstrating their utility in extracting nuanced insights that are crucial for aligning AI systems reliably in multi-cultural contexts. We show that our approach can inform rater selection and feedback interpretation by capturing nuanced viewpoints across different demographic groups, hence improving the quality of pluralistic data collection and in turn contributing to more robust AI alignment.

As large language models (LLMs) are increasingly deployed to users around the world, they are integrated into everyday tasks across diverse cultural contexts, from drafting personal communications to brainstorming creative ideas. These tasks are inherently cultural: they require contextual appropriateness, symbolic resonance, and tacit cultural expectations that native speakers draw on instinctively, meaning that a response can be factually plausible yet unmistakably wrong to a local reader. Existing cultural benchmarks have treated culture as a flat set of facts via fact verification or norm entailment methods, and have adopted LLM-as-a-Judge without examining whether they can capture such thick cultural errors. To address this gap, we present JuICE (Benchmark for LLM-Judge in Identifying Cultural Errors), a multilingual dataset of 7,470 span-level annotations of cultural and linguistic errors in long-form LLM responses. It covers 1,050 query-response pairs from four countries (the United States, South Korea, Indonesia, and Bangladesh), in both English and their countries' main languages. Using JuICE, we find that even the strongest LLM-judge achieves only an F1 of 0.52 in the erroneous span detection task. Furthermore, LLM-judges consistently miss thick cultural errors that local residents readily identify. Our findings suggest that robust cultural evaluation must move beyond surface-level detection toward frameworks that account for the depth and situatedness of cultural meaning.

The expanding capacity of AI agents for autonomous reasoning and action introduces novel discretionary domains where their decisions may carry significant consequences for individuals, communities, and society. To ensure that agents align to human expectation in near-horizon applications, they must be anchored to robust norms and principles. Currently, forward alignment (the alignment of models during training and prior to deployment (1)) often relies on under-specified reward rules that orient around mechanisms of harm rather than resultant human impact. In this position paper, we argue that meaningful operationalization of human rights during forward alignment is necessary in order to navigate novel areas of discretion on the horizon. We contend that human rights provide a core normative framework for alignment due to their human-centrism, normative specificity, global legitimacy and cultural adaptability. While human rights are occasionally referenced in alignment targets today, they are rarely used in a meaningful way to improve models during training. We validate our position by sketching a basic alignment target based on human rights and evaluating a set of 100 simulated near-future agent failures against this target. Our results illustrate that a human rights-based alignment target facilitates uniquely human-centric and normatively specific signals for model training. Finally, we describe a research agenda to advance human rights-based forward alignment in agents.

As generative AI technologies are increasingly being launched across the globe, assessing their competence to operate in different cultural contexts is exigently becoming a priority. While recent years have seen numerous and much-needed efforts on cultural benchmarking, these efforts have largely focused on specific aspects of culture and evaluation. While these efforts contribute to our understanding of cultural competence, a unified and systematic evaluation approach is needed for us as a field to comprehensively assess diverse cultural dimensions at scale. Drawing on measurement theory, we present a principled framework to aggregate multifaceted indicators of cultural capabilities into a unified assessment of cultural intelligence. We start by developing a working definition of culture that includes identifying core domains of culture. We then introduce a broad-purpose, systematic, and extensible framework for assessing cultural intelligence of AI systems. Drawing on theoretical framing from psychometric measurement validity theory, we decouple the background concept (i.e., cultural intelligence) from its operationalization via measurement. We conceptualize cultural intelligence as a suite of core capabilities spanning diverse domains, which we then operationalize through a set of indicators designed for reliable measurement. Finally, we identify the considerations, challenges, and research pathways to meaningfully measure these indicators, specifically focusing on data collection, probing strategies, and evaluation metrics.

Generative AI models ought to be useful and safe across cross-cultural contexts. One critical step toward this goal is understanding how AI models adhere to sociocultural norms. While this challenge has gained attention in NLP, existing work lacks both nuance and coverage in understanding and evaluating models' norm adherence. We address these gaps by introducing a taxonomy of norms that clarifies their contexts (e.g., distinguishing between human-human norms that models should recognize and human-AI interactional norms that apply to the human-AI interaction itself), specifications (e.g., relevant domains), and mechanisms (e.g., modes of enforcement). We demonstrate how our taxonomy can be operationalized to automatically evaluate models' norm adherence in naturalistic, open-ended settings. Our exploratory analyses suggest that state-of-the-art models frequently violate norms, though violation rates vary by model, interactional context, and country. We further show that violation rates also vary by prompt intent and situational framing. Our taxonomy and demonstrative evaluation pipeline enable nuanced, context-sensitive evaluation of cultural norm adherence in realistic settings.

The evaluation of societal biases in NLP models is critically hindered by a geo-cultural gap, This leaves regions such as Latin America severely underserved, making it impossible to adequately assess or mitigate the perpetuation of harmful regional stereotypes in language technologies. This paper presents LACES, a stereotype association dataset, for 15 Latin American countries. This dataset includes 4,789 stereotype associations manually created and annotated by 83 participants. The dataset was developed through targeted community partnerships across Latin America. Additionally, in this paper, we propose a novel adaptive data collection methodology that uniquely integrates the sourcing of new stereotype entries and the validation of existing data within a single, unified workflow. This approach results in a resource with more unique stereotypes than previous static collection methods, enabling a more efficient stereotype collection. The paper further supports the quality of LACES by demonstrating reduced efficacy of debiasing methods on this dataset in comparison to existing popular stereotype benchmarks.

Stereotype repositories are critical to assess generative AI model safety, but currently lack adequate global coverage. It is imperative to prioritize targeted expansion, strategically addressing existing deficits, over merely increasing data volume. This work introduces a multilingual stereotype resource covering four sub-Saharan African countries that are severely underrepresented in NLP resources: Ghana, Kenya, Nigeria, and South Africa. By utilizing socioculturally-situated, community-engaged methods, including telephonic surveys moderated in native languages, we establish a reproducible methodology that is sensitive to the region's complex linguistic diversity and traditional orality. By deliberately balancing the sample across diverse ethnic and demographic backgrounds, we ensure broad coverage, resulting in a dataset of 3,534 stereotypes in English and 3,206 stereotypes across 15 native languages.

Generative models are known to have reduced performance in different global cultural contexts and languages. While continual data updates have been commonly conducted to improve overall model performance, bolstering and evaluating this cross-cultural competence of generative AI models requires data resources to be intentionally expanded to include global contexts and languages. In this work, we construct a repeatable, scalable, multi-pronged pipeline to collect and contribute culturally salient, multilingual data. We posit that such data can assess the state of the global applicability of our models and thus, in turn, help identify and improve upon cross-cultural gaps.

Traffic stops are common and consequential for citizens' legal socialization and for racial gaps in police-community trust. Efforts to change the tenor of police interactions, however, may discount the discretionary context of stops-the degree of choice in the decision to stop a driver-and how organizations influence those circumstances. Discretionary stops entail more choice and thus create more ambiguity for the driver regarding the officer's intent. We examine racial disparities in the discretionary context of traffic stops, their disparate influence on community members' impressions, and the power of departmental guidance to close them. We find that stops of Black (vs. White) drivers are more likely to be for high-discretion equipment violations, and Black community members evaluate recordings of high-discretion, but not low-discretion, stops more negatively than White participants. At the same time, we find promising evidence that organizational directives to curtail equipment stops can reduce this disparity.

Ensuring the safety of Generative AI requires a nuanced understanding of pluralistic viewpoints. In this paper, we introduce a novel data-driven approach for interpreting granular ratings in pluralistic datasets. Specifically, we address the challenge of analyzing nuanced differences in safety feedback from a diverse population expressed via ordinal scales (e.g., a Likert scale). We distill non-parametric responsiveness metrics that quantify the consistency of raters in scoring varying levels of the severity of safety violations. Leveraging a publicly available pluralistic dataset of safety feedback on AI-generated content as our case study, we investigate how raters from different demographic groups (age, gender, ethnicity) use an ordinal scale to express their perceptions of the severity of violations. We apply our metrics across violation types, demonstrating their utility in extracting nuanced insights that are crucial for aligning AI systems reliably in multi-cultural contexts. We show that our approach can inform rater selection and feedback interpretation by capturing nuanced viewpoints across different demographic groups, hence improving the quality of pluralistic data collection and in turn contributing to more robust AI development.

Over a billion users across the globe interact with AI systems engineered with increasing sophistication to mimic human traits. This shift has triggered urgent debate regarding Anthropomorphism, the attribution of human characteristics to synthetic agents, and its potential to induce misplaced trust or emotional dependency. However, the causal link between more humanlike AI design and subsequent effects on engagement and trust has not been tested in realistic human-AI interactions with a global user pool. Prevailing safety frameworks continue to rely on theoretical assumptions derived from Western populations, overlooking the global diversity of AI users. Here, we address these gaps through two large-scale cross-national experiments (N=3,500) across 10 diverse nations, involving real-time and open-ended interactions with an AI system. We find that when evaluating an AI's human-likeness, users focus less on the kind of theoretical aspects often cited in policy (e.g., sentience or consciousness), but rather applied, interactional cues like conversation flow or understanding the user's perspective. We also experimentally demonstrate that humanlike design levers can causally increase anthropomorphism among users; however, we do not find that humanlike design universally increases behavioral measures for user engagement and trust, as previous theoretical work suggests. Instead, part of the connection between human-likeness and behavioral outcomes is fractured by culture: specific design choices that foster self-reported trust in AI-systems in some populations (e.g., Brazil) may trigger the opposite result in others (e.g., Japan). Our findings challenge prevailing narratives of inherent risk in humanlike AI design. Instead, we identify a nuanced, culturally mediated landscape of human-AI interaction, which demands that we move beyond a one-size-fits-all approach in AI governance.

Large language models are increasingly being integrated into applications that shape the production and discovery of societal knowledge such as search, online education, and travel planning. As a result, language models will shape how people learn about, perceive and interact with global cultures making it important to consider whose knowledge systems and perspectives are represented in models. Recognizing this importance, increasingly work in Machine Learning and NLP has focused on evaluating gaps in global cultural representational distribution within outputs. However, more work is needed on developing benchmarks for cross-cultural impacts of language models that stem from a nuanced sociologically-aware conceptualization of cultural impact or harm. We join this line of work arguing for the need of metricizable evaluations of language technologies that interrogate and account for historical power inequities and differential impacts of representation on global cultures, particularly for cultures already under-represented in the digital corpora. We look at two concepts of erasure: omission: where cultures are not represented at all and simplification i.e. when cultural complexity is erased by presenting one-dimensional views of a rich culture. The former focuses on whether something is represented, and the latter on how it is represented. We focus our analysis on two task contexts with the potential to influence global cultural production. First, we probe representations that a language model produces about different places around the world when asked to describe these contexts. Second, we analyze the cultures represented in the travel recommendations produced by a set of language model applications. Our study shows ways in which the NLP community and application developers can begin to operationalize complex socio-cultural considerations into standard evaluations and benchmarks.

The growing ubiquity of conversational AI highlights the need for frameworks that capture not only users' instrumental goals but also the situated, adaptive, and social practices through which they achieve them. Existing taxonomies of conversational behavior either overgeneralize, remain domain-specific, or reduce interactions to narrow dialogue functions. To address this gap, we introduce the Taxonomy of User Needs and Actions (TUNA), an empirically grounded framework developed through iterative qualitative analysis of 1193 human-AI conversations, supplemented by theoretical review and validation across diverse contexts. TUNA organizes user actions into a three-level hierarchy encompassing behaviors associated with information seeking, synthesis, procedural guidance, content creation, social interaction, and meta-conversation. By centering user agency and appropriation practices, TUNA enables multi-scale evaluation, supports policy harmonization across products, and provides a backbone for layering domain-specific taxonomies. This work contributes a systematic vocabulary for describing AI use, advancing both scholarly understanding and practical design of safer, more responsive, and more accountable conversational systems.

Societal stereotypes are at the center of a myriad of responsible AI interventions targeted at reducing the generation and propagation of potentially harmful outcomes. While these efforts are much needed, they tend to be fragmented and often address different parts of the issue without adopting a unified or holistic approach to social stereotypes and how they impact various parts of the machine learning pipeline. As a result, current interventions fail to capitalize on the underlying mechanisms that are common across different types of stereotypes, and to anchor on particular aspects that are relevant in certain cases. In this paper, we draw on social psychological research and build on NLP data and methods, to propose a unified framework to operationalize stereotypes in generative AI evaluations. Our framework identifies key components of stereotypes that are crucial in AI evaluation, including the target group, associated attribute, relationship characteristics, perceiving group, and context. We also provide considerations and recommendations for its responsible use.

Understanding what constitutes safety in AI-generated content is complex. While developers often rely on predefined taxonomies, real-world safety judgments also involve personal, social, and cultural perceptions of harm. This paper examines how annotators evaluate the safety of AI-generated images, focusing on the qualitative reasoning behind their judgments. Analyzing 5,372 open-ended comments, we find that annotators consistently invoke moral, emotional, and contextual reasoning that extends beyond structured safety categories. Many reflect on potential harm to others more than to themselves, grounding their judgments in lived experience, collective risk, and sociocultural awareness. Beyond individual perceptions, we also find that the structure of the task itself – including annotation guidelines – shapes how annotators interpret and express harm. Guidelines influence not only which images are flagged, but also the moral judgment behind the justifications. Annotators frequently cite factors such as image quality, visual distortion, and mismatches between prompt and output as contributing to perceived harm dimensions, which are often overlooked in standard evaluation frameworks. Our findings reveal that existing safety pipelines miss critical forms of reasoning that annotators bring to the task. We argue for evaluation designs that scaffold moral reflection, differentiate types of harm, and make space for subjective, context-sensitive interpretations of AI-generated content.

Generative large language models (LLMs) have demonstrated gaps in diverse cultural awareness across the globe. We investigate the effect of retrieval augmented generation and search-grounding techniques on LLMs' ability to display familiarity with various national cultures. Specifically, we compare the performance of standard LLMs, LLMs augmented with retrievals from a bespoke knowledge base (i.e., KB grounding), and LLMs augmented with retrievals from a web search (i.e., search grounding) on multiple cultural awareness benchmarks. We find that search grounding significantly improves the LLM performance on multiple-choice benchmarks that test propositional knowledge (e.g., cultural norms, artifacts, and institutions), while KB grounding's effectiveness is limited by inadequate knowledge base coverage and a suboptimal retriever. However, search grounding also increases the risk of stereotypical judgments by language models and fails to improve evaluators' judgments of cultural familiarity in a human evaluation with adequate statistical power. These results highlight the distinction between propositional cultural knowledge and open-ended cultural fluency when it comes to evaluating LLMs' cultural awareness.

Current text-to-image (T2I) models often fail to account for diverse human experiences, leading to misaligned systems. We advocate for pluralistic alignment, where an AI understands and is steerable towards diverse, and often conflicting, human values. Our work provides three core contributions to achieve this in T2I models. First, we introduce a novel dataset for Diverse Intersectional Visual Evaluation (DIVE) – the first multimodal dataset for pluralistic alignment. It enable deep alignment to diverse safety perspectives through a large pool of demographically intersectional human raters who provided extensive feedback across 1000 prompts, with high replication, capturing nuanced safety perceptions. Second, we empirically confirm demographics as a crucial proxy for diverse viewpoints in this domain, revealing significant, context-dependent differences in harm perception that diverge from conventional evaluations. Finally, we discuss implications for building aligned T2I models, including efficient data collection strategies, LLM judgment capabilities, and model steerability towards diverse perspectives. This research offers foundational tools for more equitable and aligned T2I systems. Content Warning: The paper includes sensitive content that may be harmful.

Human annotation plays a core role in machine learning -- annotations for supervised models, safety guardrails for generative models, and human feedback for reinforcement learning, to cite a few avenues. However, the fact that many of these human annotations are inherently subjective is often overlooked. Recent work has demonstrated that ignoring rater subjectivity (typically resulting in rater disagreement) is problematic within specific tasks and for specific subgroups. Generalizable methods to harness rater disagreement and thus understand the socio-cultural leanings of subjective tasks remain elusive. In this paper, we propose GRASP, a comprehensive disagreement analysis framework to measure group association in perspectives among different rater sub-groups, and demonstrate its utility in assessing the extent of systematic disagreements in two datasets: (1) safety annotations of human-chatbot conversations, and (2) offensiveness annotations of social media posts, both annotated by diverse rater pools across different socio-demographic axes. Our framework (based on disagreement metrics) reveals specific rater groups that have significantly different perspectives than others on certain tasks, and helps identify demographic axes that are crucial to consider in specific task contexts.

LLMs are increasingly transforming our digital ecosystem, but they often inherit societal biases learned from their training data, for instance stereotypes associating certain attributes with specific identity groups. While whether and how these biases are mitigated may depend on the specific use cases, being able to effectively detect instances of stereotype perpetuation is a crucial first step. Current methods to assess presence of stereotypes in generated language rely on simple template or co-occurrence based measures, without accounting for the variety of sentential contexts they manifest in. We argue that understanding the sentential context is crucial for detecting instances of generalization. We distinguish two types of generalizations: (1) language that merely mentions the presence of a generalization ("people think the French are very rude"), and (2) language that reinforces such a generalization ("as French they must be rude"), from non-generalizing context ("My French friends think I am rude"). For meaningful stereotype evaluations, we need to reliably distinguish such instances of generalizations. We introduce the new task of detecting generalization in language, and build GeniL, a multilingual dataset of over 50K sentences from 9 languages (English, Arabic, Bengali, Spanish, French, Hindi, Indonesian, Malay, and Portuguese) annotated for instances of generalizations. We demonstrate that the likelihood of a co-occurrence being an instance of generalization is usually low, and varies across different languages, identity groups, and attributes. We build classifiers to detect generalization in language with an overall PR-AUC of 58.7, with varying degrees of performance across languages. Our research provides data and tools to enable a nuanced understanding of stereotype perpetuation, a crucial step towards more inclusive and responsible language technologies.