Search for: All records

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. 

AI-driven tools are increasingly deployed to support low-skilled community health workers (CHWs) in hard-to-reach communities in the Global South. This paper examines how CHWs in rural India engage with and perceive AI explanations and how we might design explainable AI (XAI) interfaces that are more understandable to them. We conducted semi-structured interviews with CHWs who interacted with a design probe to predict neonatal jaundice in which AI recommendations are accompanied by explanations. We (1) identify how CHWs interpreted AI predictions and the associated explanations, (2) unpack the benefits and pitfalls they perceived of the explanations, and (3) detail how different design elements of the explanations impacted their AI understanding. Our findings demonstrate that while CHWs struggled to understand the AI explanations, they nevertheless expressed a strong preference for the explanations to be integrated into AI-driven tools and perceived several benefits of the explanations, such as helping CHWs learn new skills and improved patient trust in AI tools and in CHWs. We conclude by discussing what elements of AI need to be made explainable to novice AI users like CHWs and outline concrete design recommendations to improve the utility of XAI for novice AI users in non-Western contexts. 

Learning representations of entity mentions is a core component of modern entity linking systems for both candidate generation and making linking predictions. In this paper, we present and empirically analyze a novel training approach for learning mention and entity representations that is based on building minimum spanning arborescences (i.e., directed spanning trees) over mentions and entities across documents to explicitly model mention coreference relationships. We demonstrate the efficacy of our approach by showing significant improvements in both candidate generation recall and linking accuracy on the Zero-Shot Entity Linking dataset and MedMentions, the largest publicly available biomedical dataset. In addition, we show that our improvements in candidate generation yield higher quality re-ranking models downstream, setting a new SOTA result in linking accuracy on MedMentions. Finally, we demonstrate that our improved mention representations are also effective for the discovery of new entities via cross-document coreference.