William Brannon

profile.png

I’m a PhD student at MIT’s Center for Constructive Communication and Media Lab, advised by Deb Roy. During the PhD, I’ve interned with Brian Thompson at Amazon AI. I’m also one of the leads on the Data Provenance Initiative, an ongoing effort to audit and analyze model training datasets. Previously, I earned an MS at the Media Lab and a bachelor’s degree in mathematics from the College of William and Mary. Between undergrad and graduate school, I worked as a data scientist for various political groups and campaigns, on topics including predictive models of donor and voter behavior, design and implementation of field experiments, and data infra engineering.

I’m interested in several topics related to socially aware AI and applications to computational social science. Currently, I’m interested in and working on:

  • Data-centric AI: large language models, the role of training data, and evaluation.
  • AI for computational social science: use of machine learning and LLMs to model and understand persuasion, opinion change and media ecosystems.
  • Socially aware AI: graph deep learning and models for social network settings, especially for text-attributed graphs (TAGs).

If you’re interested in these areas and want to chat or collaborate, get in touch! The best way to reach me is by email: will.brannon@gmail.com or wbrannon@mit.edu.

For more info, you can also download my CV or check out my publications.

news

May 13, 2025 I successfully defended my PhD dissertation, “Language Models as Opinion Models: Techniques and Applications,” earlier today! The dissertation is not available online yet, but it and preprint versions of new work in it will be shortly.
Apr 26, 2025 Our paper “Bridging the Data Provenance Gap Across Text, Speech and Video” appeared today at ICLR 2025!
Jan 22, 2025 ICLR 2025 has accepted our new paper “Bridging the Data Provenance Gap Across Text, Speech and Video”! This paper is the third phase of work in the Data Provenance Initiative.
Dec 11, 2024 The latest Data Provenance Initiative paper, “Consent in Crisis: The Rapid Decline of the AI Data Commons”, appeared today at NeurIPS 2024.
Nov 12, 2024 Our paper “On the Relationship between Truth and Political Bias in Language Models” appeared as a main-conference poster at EMNLP 2024!

selected publications

  1. Shayne Longpre, Robert Mahari, Anthony Chen, Naana Obeng-Marnu, Damien Sileo, William Brannon, Niklas Muennighoff, Nathan Khazam, Jad Kabbara, Kartik Perisetla, Xinyi Wu, Enrico Shippole, Kurt Bollacker, Tongshuang Wu, Luis Villa, Sandy Pentland, and Sara Hooker
    Nature Machine Intelligence, Aug 2024
  2. William Brannon, Suyash Fulay, Hang Jiang, Wonjune Kang, Brandon Roy, Deb Roy, and Jad Kabbara
    TextGraphs at ACL, Aug 2024
  3. William Brannon and Deb Roy
    Scientific Reports, May 2024