Maarten Sap Twitter Profile Google Scholar Profile Semantic Scholar Profile

Maarten Sap

I am an assistant professor at CMU's LTI department with a courtesy appointment in HCII, and a part-time research scientist and AI safety lead at the Allen Institute for AI (AI2). My research focuses on (1) measuring and improving AI systems' social and interactional intelligence, (2) assessing and combatting social inequality, safety risks, and socio-cultural biases in human- or AI-generated language, and (3) building narrative language technologies for prosocial outcomes. I was named a 2025 Packard Fellow and a recipient of the 2025 Okawa Research Award.

I received my PhD from the University of Washington where I was advised by Noah Smith and Yejin Choi.
[bio for talks]

Recent updates:

December 2025 🏅📃: Very excited to have our paper Artificial Hivemind: The Open-Ended Homogeneity of Language Models (and Beyond) selected for a Best Paper Award at NeurIPS 2025 (Datasets and Benchmarks Track)!! Huge congrats to the first author Liwei Jiang!!!

November 2025 💎🚀: Honored to be a Spring 2025 recipient of the Amazon Research Award for our project on measuring AI agentic safety!

October 2025 🏅⭐: I’m super excited and grateful to announce that I'm part of the 2025 class of Packard Fellows. The Packard Foundation and this fellowship will allow me to explore exciting research directions towards culturally responsible and safe AI 🌍🌈

October 2025 🔍🧑‍🎓: Due to my lab being quite full already, I'm not taking looking for any new students in this upcoming PhD application cycle 😟.

October 2025 🇨🇦🎉: Excited to be attending COLM 2025 in Montreal this October! I'll be giving a talk at the Social Sim Workshop on Unlocking Social Intelligence in AI agents. I'm also thrilled that five papers I co-authored will be presented by my amazing collaborators at COLM: HAICOSYSTEM: An Ecosystem for Sandboxing Safety Risks in Human-AI Interactions (led by Xuhui Zhou et al.), ALFA: Aligning LLMs to Ask Good Questions: A Case Study in Clinical Reasoning (co-led by Jimin Mun et al.), PolyGuard: A Multilingual Safety Moderation Tool for 17 Languages, Fluid Language Model Benchmarking, and The Delta Learning Hypothesis: Preference Tuning on Weak Data can Yield Strong Gains.

August 2025 🌟: Incredibly honored to be one of 7 US recipients of the 2025 Okawa Research Grant from the Okawa Foundation!

August 2025 🧑‍🎓: Welcoming my first postdoc, Vasudha Varadarajan, to the lab!

[older news]

My research group:

Dan Chechelnitsky

CMU Portugal LTI PhD student
co-advised with Chrysoula Zerva

Joel Mire

LTI PhD student

Karina Halevy

LTI PhD student
co-advised with Mona Diab

Malia Morgan

Pre-doctoral Young Investigator at Ai2

Jimin Mun

LTI PhD student

Jocelyn Shen

MIT PhD student
co-advised with Cynthia Breazeal

Kynnedy Smith

HCII PhD student
co-advised with Motahhare Eslami

Vasudha Varadarajan

LTI Postdoc

Akhila Yerukola

LTI PhD student

Mingqian Zheng

LTI PhD student
co-advised with Carolyn Rosé

Xuhui Zhou

LTI PhD student

Overarching Research Themes

Themes extracted and images generated with the OpenAI API; there may be inconsistencies.

Human-Centered AI Safety

My research group explores how to make AI systems safer, more trustworthy, and more responsive to people in difficult or high-stakes interactions. A key focus is what happens when users are distressed or holding delusional beliefs, as in [Lost in Delusion: Examining LLM Safety Under User Delusions and Distress](https://arxiv.org/abs/2606.00975), which probes whether safety behavior remains effective under psychologically complex conditions. We also study how guardrails affect user experience and perceived usefulness, building on [Useless but Safe? Benchmarking Utility Recovery with User Intent Clarification in Multi-Turn Conversations](https://arxiv.org/abs/2604.27093) and [Let Them Down Easy! Contextual Effects of LLM Guardrails on User Perceptions and Preferences](https://arxiv.org/abs/2506.00195). Together, these papers reflect a broader effort to design AI that is not only policy-compliant, but also aligned with human needs, expectations, and vulnerability.

Narrative and Story Understanding

My research group explores how LLMs and related models understand, generate, and interpret stories, personal narratives, and narrative intent. Recent work such as [Social Story Frames: Contextual Reasoning about Narrative Intent and Reception](https://arxiv.org/abs/2512.15925) examines how context shapes what a story is trying to do and how audiences receive it. We also draw on [Words Like Knives: Backstory-Personalized Modeling and Detection of Violent Communication](https://arxiv.org/abs/2505.21451), which shows that narrative background can change how harmful language is interpreted, and [HEART-felt Narratives: Tracing Empathy and Narrative Style in Personal Stories with LLMs](https://arxiv.org/abs/2405.17633), which connects story style with empathy-related judgments. Overall, this line of work treats narrative as a rich social artifact rather than just text to classify or summarize.

Social Intelligence and Agent Simulations

My research group explores how AI agents model people, coordinate in social settings, and simulate human behavior realistically. A central theme is evaluating theory of mind and information management in interactive settings, as seen in [SOTOPIA-ToM: Evaluating Information Management in Multi-Agent Interaction with Theory of Mind](https://arxiv.org/abs/2605.02307) and [SoMi-ToM: Evaluating Multi-Perspective Theory of Mind in Embodied Social Interactions](https://arxiv.org/abs/2506.23046). We also study the limits of social simulation itself through [Reinforcing Human Behavior Simulation via Verbal Feedback](https://arxiv.org/abs/2605.20506) and [Mind the Sim2Real Gap in User Simulation for Agentic Tasks](https://arxiv.org/abs/2603.11245), which highlight how simulated users can diverge from real people. This research helps clarify when agents are socially intelligent, when they merely appear that way, and how to benchmark those differences.

Bias, Culture, and Responsible Language Models

My research group explores how language models reflect, amplify, or mitigate bias across languages, cultures, and interaction styles. We focus on how models behave under cultural variation and socially grounded norms, including [NormAd: A Framework for Measuring the Cultural Adaptability of Large Language Models](https://aclanthology.org/2025.naacl-long.120/) and [Mind the Gesture: Evaluating AI Sensitivity to Culturally Offensive Non-Verbal Gestures](https://arxiv.org/abs/2502.17710). Another important direction is fairness in representation and moderation, as shown by [Rejected Dialects: Biases Against African American Language in Reward Models](https://arxiv.org/abs/2502.12858) and [PolyGuard: A Multilingual Safety Moderation Tool for 17 Languages](https://arxiv.org/abs/2504.04377). Together, these studies point toward more culturally aware, multilingual, and socially responsible AI systems.

More about me