Jay Chooi

I work on preparing society for powerful artificial intelligence.

Currently, I am a Research Fellow at MATS. My research interests are forecasting physical automation, and understanding how society can remain stable and flourish when the political power and economic leverage from human labor diminish.

I graduated from Harvard with a Master's in Statistics and a Bachelor's in Computer Science and Mathematics. There, I was fortunate to be mentored by Professor Ariel Procaccia. My thesis, Computational Perspectives on Democracy in the Age of AI, was awarded Highest Honors.

I am a Rhodes Scholar for 2026. In 2024, my US Presidential Election forecast got all 50 states correct. I am also a gold medalist at the 2021 International Olympiad in Astronomy and Astrophysics.

Read more about my motivations.

Working Papers

Efficient Ensemble Selection from Binary and Pairwise Feedback. [arxiv] with T. Y. Neoh, N. Teh, P. W. Goldberg, M. Tambe

Covert Influence Between Language Models. [commentable] with A. Shah, S. Feng

Bridging in Participatory Budgeting. [commentable] with T. Y. Neoh, A. Satish, A. Laprevotte

A Universal Measure of Labour Market Disruption: A Case Study on the UK, 1960–2020. [commentable]

The Persistent Non-Response Bias of Trump Voters in a Sample-Matched Poll. [commentable]

Research

Finding Common Ground in a Sea of Alternatives. [arxiv] ACM Conference on Economics and Computation (EC 2026) with P. Gölz, A. Procaccia, B. Schiffer, S. Zhang

05/2026

Measuring AI-Induced Disempowerment: A Framework and Proposed Metrics. [commentable] ACL 2026 Workshop on Evaluating AI in Practice (Eval Eval) with J. Lee, J. Li

04/2026

Stylistic Transfer from Annotator Communities to Large Language Models. [official] ACL 2026 Joint SIGHUM Workshop on Computational Linguistics for Cultural Heritage, Social Sciences, Humanities and Literature (LaTeCH-CLfL)

01/2026

DP-AdamW: Investigating Decoupled Weight Decay and Bias Correction in Private Deep Learning. [arxiv] [official] ICML 2025 Workshop on Collaborative and Federated Agentic Workflows with K. Cong, R. Li, L. Sun

11/2025

BluePrint: A Social Media User Dataset for LLM Persona Evaluation and Training. [arxiv] EMNLP 2025 Workshop on Tailoring AI: Exploring Active and Passive LLM Personalization (PALS) with A. Bück-Kaeffer, D. Zhao, M. Puelma Touzel, K. Pelrine, J.-F. Godbout, R. Rabbany, Z. Yang

10/2025

OpenAI Is Huge in India. Its Models Are Steeped in Caste Bias. [official] MIT Technology Review with N. Christopher

10/2025

Understanding the Landscape of Agrifood Smallholders in Malaysia: Climate Risks, Sustainable Standards, and Gender Gap. [official] Khazanah Research Institute with S. Che Omar, M. A. R. Abu Rahim, N. S. A. Nik Sharifulden, A. N. Teoh, C. X. Lai

03/2024