I'm an Assistant Professor of Applied AI and Kathryn and Grant Swick Faculty Scholar at UChicago Booth, where I work on behavior-bound machine learning.

Machine learning is not a sterile industrial process; much in the way that it is hardware-bound and software-bound, it is also bound by the behavior of real-world actors such as workers, firms, and states. I create machines that are compatible with actual actors, not just idealized ones, with research across post-training, evaluation, and AI governance.

My work has received an ICML 2022 Outstanding Paper award, a Facebook Fellowship, and an NSERC PGS-D. Prior to UChicago, I graduated with a PhD in Computer Science from Stanford University, where I was advised by Dan Jurafsky. Notable work:

• • How Contextual are Contextualized Word Representations? Comparing the Geometry of BERT, ELMo, and GPT-2 Embeddings.
    Kawin Ethayarajh.
    EMNLP 2019.
    paper tweet code
  Contextual representations in LLMs are anisotropic: they occupy a narrow cone in vector space instead of being uniformly distributed. This paper identified and named the phenomenon, which has since become a foundational concept in AI—cited as core motivation for contrastive learning approaches like SimCSE, embedding post-processing methods like whitening, and a broad literature on representations.
• • Understanding Dataset Difficulty with V-Usable Information.
    Kawin Ethayarajh, Yejin Choi, and Swabha Swayamdipta.
    ICML 2022 (outstanding paper - top 10 of 5,630 submissions).
    paper tweet 1 tweet 2 code dataset 1 dataset 2
  We propose framing dataset difficulty as the lack of V-usable information. This led to the development of Stanford Human Preferences (SHP), the first large-scale dataset of human preferences over natural text, based on data from Reddit. SHP was the only dataset from academia used to post-train Llama-2, one of the first open-weight models to be preference-tuned and one of the most downloaded LLMs ever. Within a year of SHP's release, Reddit started licensing its data for AI training in deals valued at 200M+ USD.
• • Model Alignment as Prospect Theoretic Optimization.
    Kawin Ethayarajh, Winnie Xu, Niklas Muennighoff, Dan Jurafsky, and Douwe Kiela.
    ICML 2024 (spotlight - top 335 of 9,473 submissions).
    paper tweet code press
  We prove that the top post-training objectives (PPO, DPO) capture perceptual biases in how humans see random variables. Surprisingly, these biases—like loss aversion—make post-training more efficient and performant than it otherwise would be. By more intentionally integrating these biases, we create Kahneman-Tversky Optimization (KTO), a widely-used post-training algorithm..
• • Mecha-nudges for Machines.
    Giulio Frey and Kawin Ethayarajh.
    Technical AI Governance Workshop (TAIGR) @ ICML 2026 (outstanding paper - top 6 of 174 submissions).
    Game Theory and Mechanism Design Workshop @ EC 2026 (spotlight)
    paper tweet code
  As AI agents make decisions in the same environments as humans, the environments themselves will change to influence them. We formalize these as mecha-nudges and provide the first large-scale evidence that systematic mecha-nudging is already happening in the wild, by studying the Etsy marketplace.