Aditya Vashistha

Dhruv Agarwal, Mor Naaman, Aditya Vashistha

Large language models (LLMs) are being increasingly integrated into everyday products and services, such as coding tools and writing assistants. As these embedded AI applications are deployed globally, there is a growing concern that the AI models underlying these applications prioritize Western values. This paper investigates what happens when a Western-centric AI model provides writing suggestions to users from a different cultural background. We conducted a cross-cultural controlled experiment with 118 participants from India and the United States who completed culturally grounded writing tasks with and without AI suggestions. Our analysis reveals that AI provided greater efficiency gains for Americans compared to Indians. Moreover, AI suggestions led Indian participants to adopt Western writing styles, altering not just what is written but also how it is written. These findings show that Western-centric AI models homogenize writing toward Western norms, diminishing nuances that differentiate cultural expression.

Cody Kommers, Ruth Ahnert, Maria Antoniak et al.

Generative AI systems are increasingly recognized as cultural technologies, yet current evaluation frameworks often treat culture as a variable to be measured rather than fundamental to the system's operation. Drawing on hermeneutic theory from the humanities, we argue that GenAI systems function as "context machines" that must inherently address three interpretive challenges: situatedness (meaning only emerges in context), plurality (multiple valid interpretations coexist), and ambiguity (interpretations naturally conflict). We present computational hermeneutics as an emerging framework offering an interpretive account of what GenAI systems do, and how they might do it better. We offer three principles for hermeneutic evaluation -- that benchmarks should be iterative, not one-off; include people, not just machines; and measure cultural context, not just model output. This perspective offers a nascent paradigm for designing and evaluating contemporary AI systems: shifting from standardized questions about accuracy to contextual ones about meaning.

Agrima Seth, Monojit Choudhary, Sunayana Sitaram et al.

Representational bias in large language models (LLMs) has predominantly been measured through single-response interactions and has focused on Global North-centric identities like race and gender. We expand on that research by conducting a systematic audit of GPT-4 Turbo to reveal how deeply encoded representational biases are and how they extend to less-explored dimensions of identity. We prompt GPT-4 Turbo to generate over 7,200 stories about significant life events (such as weddings) in India, using prompts designed to encourage diversity to varying extents. Comparing the diversity of religious and caste representation in the outputs against the actual population distribution in India as recorded in census data, we quantify the presence and "stickiness" of representational bias in the LLM for religion and caste. We find that GPT-4 responses consistently overrepresent culturally dominant groups far beyond their statistical representation, despite prompts intended to encourage representational diversity. Our findings also suggest that representational bias in LLMs has a winner-take-all quality that is more biased than the likely distribution bias in their training data, and repeated prompt-based nudges have limited and inconsistent efficacy in dislodging these biases. These results suggest that diversifying training data alone may not be sufficient to correct LLM bias, highlighting the need for more fundamental changes in model development. Dataset and Codebook: https://github.com/agrimaseth/How-Deep-Is-Representational-Bias-in-LLMs

Gauri Nayak, Farhana Shahid, Aditya Vashistha et al.

WhatsApp is one of the most widely used messaging platforms globally, with billions of users sharing information in private groups. Yet, it offers little infrastructure to support moderation and group governance. In the absence of platform-level oversight, group admins bear the responsibility of governing group behavior. In this paper, we explore how WhatsApp group admins collaborate with AI tools to create, enforce, and maintain group rules. Drawing on a two-phase speculative design study with 20 admins in India, we examine how participants interacted with an AI assistant (Meta AI) to co-create rules and responded to a series of probes illustrating AI-assisted moderation features. Our findings show that while admins appreciated the AI's ability to surface overlooked rules and reduce their moderation burden, they were highly sensitive to issues of relational trust, data privacy, tone, and social context. We identify how group type and admin style shaped their willingness to delegate authority, and surface the limitations of current chatbot interfaces in supporting collaborative rule-making. We conclude with design implications for building moderation tools that center human judgment, relational nuance, contextual adaptability, and collective governance.

Rishav Hada, Safiya Husain, Varun Gumma et al. · microsoft-research

Existing research in measuring and mitigating gender bias predominantly centers on English, overlooking the intricate challenges posed by non-English languages and the Global South. This paper presents the first comprehensive study delving into the nuanced landscape of gender bias in Hindi, the third most spoken language globally. Our study employs diverse mining techniques, computational models, field studies and sheds light on the limitations of current methodologies. Given the challenges faced with mining gender biased statements in Hindi using existing methods, we conducted field studies to bootstrap the collection of such sentences. Through field studies involving rural and low-income community women, we uncover diverse perceptions of gender bias, underscoring the necessity for context-specific approaches. This paper advocates for a community-centric research design, amplifying voices often marginalized in previous studies. Our findings not only contribute to the understanding of gender bias in Hindi but also establish a foundation for further exploration of Indic languages. By exploring the intricacies of this understudied context, we call for thoughtful engagement with gender bias, promoting inclusivity and equity in linguistic and cultural contexts beyond the Global North.

Hamna, Deepthi Sudharsan, Agrima Seth et al.

Large Language Models (LLMs) and Text-To-Image (T2I) models have demonstrated the ability to generate compelling text and visual stories. However, their outputs are predominantly aligned with the sensibilities of the Global North, often resulting in an outsider's gaze on other cultures. As a result, non-Western communities have to put extra effort into generating culturally specific stories. To address this challenge, we developed a visual storytelling tool called Kahani that generates culturally grounded visual stories for non-Western cultures. Our tool leverages off-the-shelf models GPT-4 Turbo and Stable Diffusion XL (SDXL). By using Chain of Thought (CoT) and T2I prompting techniques, we capture the cultural context from user's prompt and generate vivid descriptions of the characters and scene compositions. To evaluate the effectiveness of Kahani, we conducted a comparative user study with ChatGPT-4 (with DALL-E3) in which participants from different regions of India compared the cultural relevance of stories generated by the two tools. The results of the qualitative and quantitative analysis performed in the user study show that Kahani's visual stories are more culturally nuanced than those generated by ChatGPT-4. In 27 out of 36 comparisons, Kahani outperformed or was on par with ChatGPT-4, effectively capturing cultural nuances and incorporating more Culturally Specific Items (CSI), validating its ability to generate culturally grounded visual stories.

Dhruv Agarwal, Anya Shukla, Sunayana Sitaram et al.

Large language models (LLMs) are used worldwide, yet exhibit Western cultural tendencies. Many countries are now building ``regional'' LLMs, but it remains unclear whether they reflect local values and practices or merely speak local languages. Using India as a case study, we evaluate six Indic and six global LLMs on two dimensions -- values and practices -- grounded in nationally representative surveys and community-sourced QA datasets. Across tasks, Indic models do not align better with Indian norms than global models; in fact, a U.S. respondent is a closer proxy for Indian values than any Indic model. Prompting and regional fine-tuning fail to recover alignment and can even degrade existing knowledge. We attribute this to scarce culturally grounded data, especially for pretraining. We position cultural evaluation as a first-class requirement alongside multilingual benchmarks and offer a reusable, community-grounded methodology. We call for native, community-authored corpora and thick x wide evaluations to build truly sovereign LLMs.

Kirti Bhagat, Shaily Bhatt, Athul Velagapudi et al.

Millions of users across the globe turn to AI chatbots for their creative needs, inviting widespread interest in understanding how they represent diverse cultures. However, evaluating cultural representations in open-ended tasks remains challenging and underexplored. In this work, we present TALES, an evaluation of cultural misrepresentations in LLM-generated stories for diverse Indian cultural identities. First, we develop TALES-Tax, a taxonomy of cultural misrepresentations by collating insights from participants with lived experiences in India through focus groups (N=9) and individual surveys (N=15). Using TALES-Tax, we evaluate 6 models through a large-scale annotation study spanning 2925 annotations from 108 annotators with lived experience and native language proficiency from across 71 regions in India and 14 languages. Concerningly, we find that 88% of the generated stories contain misrepresentations, and such errors are more prevalent in mid- and low-resourced languages and stories based in peri-urban regions in India. We also transform the annotations into TALES-QA, a standalone question bank to evaluate the cultural knowledge of models.

Mahika Phutane, Hayoung Jung, Matthew Kim et al. · uw

Large language models (LLMs) are increasingly under scrutiny for perpetuating identity-based discrimination in high-stakes domains such as hiring, particularly against people with disabilities (PwD). However, existing research remains largely Western-centric, overlooking how intersecting forms of marginalization--such as gender and caste--shape experiences of PwD in the Global South. We conduct a comprehensive audit of six LLMs across 2,820 hiring scenarios spanning diverse disability, gender, nationality, and caste profiles. To capture subtle intersectional harms and biases, we introduce ABLEIST (Ableism, Inspiration, Superhumanization, and Tokenism), a set of five ableism-specific and three intersectional harm metrics grounded in disability studies literature. Our results reveal significant increases in ABLEIST harms towards disabled candidates--harms that many state-of-the-art models failed to detect. These harms were further amplified by sharp increases in intersectional harms (e.g., Tokenism) for gender and caste-marginalized disabled candidates, highlighting critical blind spots in current safety tools and the need for intersectional safety evaluations of frontier models in high-stakes domains like hiring.

Mahika Phutane, Aditya Vashistha

People with disabilities (PwD) experience disproportionately high levels of discrimination and hate online, particularly in India, where entrenched stigma and limited resources intensify these challenges. Large language models (LLMs) are increasingly used to identify and mitigate online hate, yet most research on online ableism focuses on Western audiences with Western AI models. Are these models adequately equipped to recognize ableist harm in non-Western places like India? Do localized, Indic language models perform better? To investigate, we adopted and translated a publicly available ableist speech dataset to Hindi, and prompted eight LLMs--four developed in the U.S. (GPT-4, Gemini, Claude, Llama) and four in India (Krutrim, Nanda, Gajendra, Airavata)--to score and explain ableism. In parallel, we recruited 175 PwD from both the U.S. and India to perform the same task, revealing stark differences between groups. Western LLMs consistently overestimated ableist harm, while Indic LLMs underestimated it. Even more concerning, all LLMs were more tolerant of ableism when it was expressed in Hindi and asserted Western framings of ableist harm. In contrast, Indian PwD interpreted harm through intention, relationality, and resilience--emphasizing a desire to inform and educate perpetrators. This work provides groundwork for global, inclusive standards of ableism, demonstrating the need to center local disability experiences in the design and evaluation of AI systems.

Farhana Shahid, Mona Elswah, Aditya Vashistha

Most social media users come from the Global South, where harmful content usually appears in local languages. Yet, AI-driven moderation systems struggle with low-resource languages spoken in these regions. Through semi-structured interviews with 22 AI experts working on harmful content detection in four low-resource languages: Tamil (South Asia), Swahili (East Africa), Maghrebi Arabic (North Africa), and Quechua (South America)--we examine systemic issues in building automated moderation tools for these languages. Our findings reveal that beyond data scarcity, socio-political factors such as tech companies' monopoly on user data and lack of investment in moderation for low-profit Global South markets exacerbate historic inequities. Even if more data were available, the English-centric and data-intensive design of language models and preprocessing techniques overlooks the need to design for morphologically complex, linguistically diverse, and code-mixed languages. We argue these limitations are not just technical gaps caused by "data scarcity" but reflect structural inequities, rooted in colonial suppression of non-Western languages. We discuss multi-stakeholder approaches to strengthen local research capacity, democratize data access, and support language-aware solutions to improve automated moderation for low-resource languages.