Maarten Sap

My research group explores the complexities of ethics and bias in artificial intelligence systems. A key paper, [Mind the Gesture: Evaluating AI Sensitivity to Culturally Offensive Non-Verbal Gestures](https://arxiv.org/abs/2502.17710), investigates how AI can be sensitive to cultural nuances in communication, which is crucial for developing more inclusive models. Additionally, the work on [Not Like Us, Hunty: Measuring Perceptions and Behavioral Effects of Minoritized Anthropomorphic Cues in LLMs](https://arxiv.org/abs/2505.05660) highlights the importance of understanding how anthropomorphic designs affect user perceptions across diverse demographics. Another significant study, [Particip-AI: A Democratic Surveying Framework for Anticipating Future AI Use Cases, Harms and Benefits](https://arxiv.org/abs/2403.14791), focuses on engaging the community in discussions about AI's potential impacts, ensuring that diverse voices contribute to responsible AI development.

My research group explores how narratives shape human experience and understanding, especially in the context of AI-generated content. A recent paper, [Quantifying the narrative flow of imagined versus autobiographical stories](https://www.pnas.org/doi/10.1073/pnas.2211715119), delves into the differences between personal storytelling and constructed narratives, enhancing our comprehension of narrative structures. Another important study, [HEART-felt Narratives: Tracing Empathy and Narrative Style in Personal Stories with LLMs](https://arxiv.org/abs/2405.17633), examines how AI can influence emotional engagement in storytelling. Furthermore, the work on [The Empirical Variability of Narrative Perceptions of Social Media Texts](https://aclanthology.org/2024.emnlp-main.1113/) provides insight into the diverse ways audiences interpret narratives across digital platforms.

My research group explores the intersection of social intelligence and AI agent interactions. The paper [SOTOPIA: Interactive Evaluation for Social Intelligence in Language Agents](https://arxiv.org/abs/2310.11667) sets the groundwork for assessing how well AI can navigate social situations and exhibit appropriate responses. Additionally, [AI-LieDar: Examine the Trade-off Between Utility and Truthfulness in LLM Agents](https://aclanthology.org/2025.naacl-long.595/) discusses the critical balance between functionality and ethical considerations in AI conversations. Finally, our work on [SOTOPIA-S4: A User-Friendly System for Flexible, Customizable, and Large-Scale Social Simulation](https://arxiv.org/abs/2504.16122) offers a framework for engaging users in complex social scenarios using AI.

My research group explores strategies to reduce bias and improve fairness in language models. A landmark study, [Mitigating Bias in RAG: Controlling the Embedder](https://arxiv.org/abs/2502.17390), investigates the influence of embeddings on bias propagation in generated content. Another important contribution, [Disparities in LLM Reasoning Accuracy and Explanations: A Case Study on African American English](https://arxiv.org/abs/2503.04099), emphasizes the need to evaluate reasoning capabilities across diverse linguistic contexts. Additionally, [Rejected Dialects: Biases Against African American Language in Reward Models](https://arxiv.org/abs/2502.12858) sheds light on how language models can inadvertently reinforce stereotypes, underscoring the need for systematic bias audits.