Tao Lin
Tao Lin

PhD in Computer Science

Harvard University

About Me

Starting in July 2025, I will be a postdoctoral researcher at Microsoft Research, in the Economics and Computation group in New England, hosted by Alex Slivkins.

I obtained my PhD degree in Computer Science from Harvard University, where I am very fortunate to be advised by Yiling Chen. My research lies in the intersection between economics and machine learning. I have been working on “mechanism design + machine learning” problems since my undergraduate study at Peking University, working with Xiaotie Deng. Recently, I focused more on information design problems, like Bayesian persuasion. I also investigate the incentive issues in real-world machine learning systems, such as ad auction platforms and recommender systems. From 2023 to 2024, I interned at ByteDance and Google. I received the Siebel Scholarship in 2024.

I will be an assistant professor in the School of Data Science at the Chinese University of Hong Kong, Shenzhen, starting in 2026.

Contact: tlin@g.harvard.edu

Interests
  • Information Design
  • Mechanism Design
  • Machine Learning
Education

  • PhD in Computer Science

    Harvard University

  • BSc in EECS

    Peking University

🗞️ News
📚 My Research
My research direction is Learning-Based Incentive Design. It is motivated by the question of “how to design a multi-agent system so that self-interested agents are incentivized to achieve a globally desirable outcome”. In particular, I study mechanism design and informamtion design problems with machine-learning-based decision-makers. I am also fascinated by the incentive issues in real-world machine learning systems, such as ad auction platforms and recommender systems.
📑 Featured Publications
Generalized Principal-Agent Problem with a Learning Agent

Generalized Principal-Agent Problem with a Learning Agent

Classic principal-agent problems such as Stackelberg games, contract design, and Bayesian persuasion, often assume that the agent is able to best respond to the principal’s committed strategy. We study repeated generalized principal-agent problems under the assumption that the principal does not have commitment power and the agent uses algorithms to learn to respond to the principal. We reduce this problem to a one-shot generalized principal-agent problem where the agent approximately best responds. Using this reduction, we show that: (1) If the agent uses contextual no-regret learning algorithms with regret $\mathrm{Reg}(T)$, then the principal can guarantee utility at least $U^* - \Theta\big(\sqrt{\tfrac{\mathrm{Reg}(T)}{T}}\big)$, where $U^*$ is the principal’s optimal utility in the classic model with a best-responding agent. (2) If the agent uses contextual no-swap-regret learning algorithms with swap-regret $\mathrm{SReg}(T)$, then the principal cannot obtain utility more than $U^* + O(\frac{\mathrm{SReg(T)}}{T})$. But (3) if the agent uses mean-based learning algorithms (which can be no-regret but not no-swap-regret), then the principal can sometimes do significantly better than $U^*$. These results not only refine previous results in Stackelberg games and contract design, but also lead to new results for Bayesian persuasion with a learning agent and all generalized principal-agent problems where the agent does not have private information.

User-Creator Feature Polarization in Recommender Systems with Dual Influence

User-Creator Feature Polarization in Recommender Systems with Dual Influence

Recommender systems serve the dual purpose of presenting relevant content to users and helping content creators reach their target audience. The dual nature of these systems naturally influences both users and creators: users’ preferences are affected by the items they are recommended, while creators may be incentivized to alter their content to attract more users. We define a model, called user-creator feature dynamics, to capture the dual influence of recommender systems. We prove that a recommender system with dual influence is guaranteed to polarize, causing diversity loss in the system. We then investigate, both theoretically and empirically, approaches for mitigating polarization and promoting diversity in recommender systems. Unexpectedly, we find that common diversity-promoting approaches do not work in the presence of dual influence, while relevancy-optimizing methods like top-$k$ truncation can prevent polarization and improve diversity of the system.

📃 Working Papers
Brian Hu Zhang*, Tao Lin*, Yiling Chen, Tuomas Sandholm (2025). Learning a Game by Paying the Agents. Working paper.
Cite URL
📃 Conference Publications
Tao Lin, Yiling Chen (2025). Generalized Principal-Agent Problem with a Learning Agent. International Conference on Learning Representations (ICLR spotlight).
PDF Cite Poster URL
Tao Lin, Ce Li (2025). Information Design with Unknown Prior. Proceedings of Innovations in Theoretical Computer Science (ITCS).
PDF Cite Video URL
Tao Lin, Kun Jin, Andrew Estornell, Xiaoying Zhang, Yiling Chen, Yang Liu (2024). User-Creator Feature Polarization in Recommender Systems with Dual Influence. Advances in Neural Information Processing Systems (NeurIPS).
PDF Cite Poster Slides URL
(Alphabetical) Yiling Chen, Tao Lin, Ariel D. Procaccia, Aaditya Ramdas, Itai Shapira (2024). Bias Detection via Signaling. Advances in Neural Information Processing Systems (NeurIPS).
PDF Cite URL
Safwan Hossain*, Tonghan Wang*, Tao Lin*, Yiling Chen, David C. Parkes, Haifeng Xu (2024). Multi-Sender Persuasion: A Computational Perspective. International Conference on Machine Learning (ICML).
PDF Cite URL
Jiahao Zhang, Tao Lin, Weiqiang Zheng, Zhe Feng, Yifeng Teng, Xiaotie Deng (2024). Learning Thresholds with Latent Values and Censored Feedback. International Conference on Learning Representations (ICLR).
PDF Cite URL
Tao Lin, Yiling Chen (2023). Sample Complexity of Forecast Aggregation. Advances in Neural Information Processing Systems (NeurIPS spotlight).
PDF Cite Poster arXiv video in Chinese
(Alphabetical) Xiaotie Deng, Yotam Gafni, Ron Lavi, Tao Lin, Hongyi Ling (2023). From Monopoly to Competition: Optimal Contests Prevail. Proceedings of the AAAI Conference on Artificial Intelligence.
PDF Cite Poster URL
(Alphabetical) Xiaotie Deng, Xinyan Hu, Tao Lin, Weiqiang Zheng (2022). Nash Convergence of Mean-Based Learning Algorithms in First Price Auctions. Proceedings of the ACM Web Conference (WWW).
PDF Cite Slides URL arxiv 15-min video 2-min video by Xinyan
Manon Revel, Tao Lin, Daniel Halpern (2022). How Many Representatives Do We Need? The Optimal Size of a Congress Voting on Binary Issues. Proceedings of the AAAI Conference on Artificial Intelligence.
Cite URL
(Alphabetical) Hu Fu, Tao Lin (2020). Learning Utilities and Equilibria in Non-Truthful Auctions. Advances in Neural Information Processing Systems (NeurIPS).
PDF Cite arxiv
(Alphabetical) Xiaotie Deng, Ron Lavi, Tao Lin, Qi Qi, Wenwei Wang, Xiang Yan (2020). A Game-Theoretic Analysis of the Empirical Revenue Maximization Algorithm with Endogenous Sampling. Advances in Neural Information Processing Systems (NeurIPS).
PDF Cite arXiv
(Alphabetical) Xiaotie Deng, Tao Lin, Tao Xiao (2020). Private Data Manipulation in Optimal Sponsored Search Auction. Proceedings of The Web Conference (WWW).
PDF Cite URL
📃 Journal Publications
(Alphabetical) Xiaotie Deng, Yotam Gafni, Ron Lavi, Tao Lin, Hongyi Ling (2024). From Monopoly to Competition: When Do Optimal Contests Prevail?. Minor revision at Games and Economic Behavior.
Cite URL
📝 Notes
Tao Lin (2022). How Does Independence Help Generalization? Sample Complexity of ERM on Product Distributions.
Cite URL
(Alphabetical) Xiaoming Li, Tao Lin (2019). On Clearing Prices in Matching Markets: A Simple Characterization without Duality.
Cite URL
🎙️ Recent & Upcoming Talks
See all

Experience

  1. Postdoctoral Researcher

    Microsoft Research
    Hosted by Alex Slivkins

  2. Student Researcher

    Google
    Hosted by Christopher Liaw

  3. Research Intern

    ByteDance
    Hosted by Yang Liu

Education

  1. PhD in Computer Science

    Harvard University
    Dissertation: Incentive Design in the Machine Learning Age

  2. BSc in EECS

    Peking University