The Human-Centered AI Lab is a nonprofit research organization investigating how AI systems process language, make decisions, and fail. Our work focuses on the vulnerabilities in large language models that purely technical evaluation approaches miss, particularly in how models handle negation, prohibition, persuasion, and emotional manipulation.

We are Principal Investigators in the NIST AI Safety Institute Consortium, representing the Modern Language Association, and PIs for a Schmidt Sciences HAVI grant building open-access AI tools to rescue endangered cultural archives in New Orleans. Our team spans Kenyon College, Columbia University, Berklee College of Music, and LSU.

Founded in 2016 by Katherine Elkins and Jon Chun, the lab grew out of what is believed to be the first human-centered AI curriculum in higher education. Our students and researchers have produced over 300 AI/ML research projects, downloaded 95,000+ times from institutions in 198 countries. Published work appears in Cambridge University Press, ICML, Frontiers in Computer Science, the Journal of Cultural Analytics, and Narrative, among others.

Our research spans AI safety and LLM evaluation, computational social science, sentiment analysis, archival intelligence, and AI governance. We work across federal (NIST), international (UNESCO), and industry (Meta, OpenAI, Bloomberg) contexts.

Featured in Forbes, the Christian Science Monitor, and NPR.

Human-Centered AI Lab

Org chart

Teams

Offices