Cultural AI Lab

Duke University

The Cultural AI Lab studies what happens when machines become participants in culture — making it, consuming it, judging it, and increasingly judging us.

Undergraduate and graduate students in the humanities, computer science, and data science work together to study the impact of generative AI on cultural production and consumption, and how we might better redesign these systems for humans. We combine large-scale computational analysis with experiments and surveys, and we read closely.

We currently work on three topics. We study how judgment and taste are changing in the age of AI slop, when content is abundant and discernment is scarce. We build creative AI systems around the interpretive communities that actually read, rather than a generic average reader. And we track what happens to culture as LLMs saturate platforms like Reddit and Substack. See the projects page for current studies.

The lab has been generously funded by Schmidt Sciences, through the Humanities and AI Virtual Institute, and by the Social Sciences and Humanities Research Council of Canada.

Selected work

  1. AI ‘Slop’ Is a Social Judgement
    Nature Human Behaviour, 2026. Read paper Richard Jean So, Hoyt Long, Cody Kommers, Mina Lee, Emily Wenger, Peter West, Eamon Duede, Ari Holtzman, Chenhao Tan.
  2. The Social and Cultural Life of Artificial Intelligence (AI) Slop: A Research Framework
    Big Data & Society, 2026. Read paper Richard Jean So, Hoyt Long.
  3. Critical Confabulation: Can LLMs Hallucinate for Social Good?
    International Conference on Learning Representations (ICLR), 2026. Read paper Peiqi Sui, Eamon Duede, Richard Jean So, Hoyt Long.
  4. The Social AI Author: Modeling Creativity and Distinction in Simulated Cultural Fields
    AI & Society, 2026. Read paper Edwin Roland, Richard Jean So, Hoyt Long.
  5. What Does AI Do for Cultural Interpretation? A Randomized Experiment on Close Reading with Exposure to AI
    ACM Conference on Human Factors in Computing Systems (CHI), 2026. Read paper Jiayin Zhi, Hoyt Long, Richard Jean So, Mina Lee.
  6. Spoiler Alert: Narrative Forecasting as a Metric for Tension in LLM Storytelling
    Conference on Language Modeling (COLM), 2026. Read preprint Peiqi Sui, Yutong Zhu, Tianyi Cheng, Peter West, Richard Jean So, Hoyt Long, Ari Holtzman.

Get involved

Faculty, graduate students, and undergraduates interested in this work are welcome to get in touch. Prospective PhD applicants in English should write directly.