I am an AI safety researcher and Co-Founder of the Human-Centered AI Lab, a PI representing the MLA in the NIST AI Safety Institute Consortium, and a PI for a Schmidt Sciences HAVI grant building open-access AI tools for endangered cultural archives.
My research focuses on where language models break down--how they mishandle negation, respond to emotional manipulation, and fail at the kinds of prohibition and persuasion humans navigate every day. I build computational tools to find and measure these failures.
Current work:
- LLM evaluation for the NIST AI Safety Institute Consortium (representing the Modern Language Association)
- Archival Intelligence: AI tools for endangered archives in New Orleans (Schmidt Sciences HAVI)
- SentimentArcs: computational methodology for emotional arc in narrative
- Multi-agent behavioral simulation for high-stakes decision-making
Links: Website · Google Scholar · Human-Centered AI Lab · ORCID