Why information cocoons emerge
Revealing how the adaptive feedback between human preferences and recommendation algorithms drives the loss of information diversity.
Postdoctoral Researcher · Future Intelligence Lab
Understanding human–AI systems. Building artificial societies.

I am a postdoctoral researcher in the Department of Electronic Engineering at Tsinghua University, working with Prof. Yong Li at the Future Intelligence Lab. I received my Ph.D. (2026) and M.Eng. (2022) from Tsinghua University, advised by Prof. Yong Li, and my B.Eng. from Tianjin University (2019).
My research lies at the intersection of artificial intelligence, complex systems, and computational social science. I study how people and intelligent algorithms shape one another, how collective behavior emerges in networks of LLM agents, and how we can build useful artificial societies for social science.
My work has appeared in Nature Machine Intelligence and Nature Communications. I lead the development of AgentSociety, an open research platform for large-scale social simulation and computational experiments.
Feedback loops between people and algorithms, information cocoons, and the social impact of recommendation.
Social cognition, opinion dynamics, and collective behavior in networks of intelligent agents.
Large-scale artificial societies and reproducible experiments for social and economic systems.
Revealing how the adaptive feedback between human preferences and recommendation algorithms drives the loss of information diversity.
Showing how polarization emerges in networks of LLM agents, and using these systems to investigate social mechanisms and possible interventions.
Investigating persistent semantic convergence in interacting LLM systems, even when their wording varies. Joint work with Weiyi Kong, Shiyang Lai, and James Evans.
An open platform for studying human behavior and society with LLM agents—from large-scale social simulation to an integrated workflow for computational experiments.
LLM agents interacting in realistic social environments.
AI social scientists connecting hypotheses, simulations, and analysis.
Our paper Multi-LLM Systems Exhibit Robust Semantic Collapse, with Weiyi Kong, Shiyang Lai, and James Evans, has been accepted to Nature Machine Intelligence.
Our work on polarization in networks of LLM agents is published in Nature Communications.
I joined Tsinghua University as a postdoctoral researcher in the Department of Electronic Engineering.
The AgentSociety paper is available online in iFuture.
MarketSim, our work on large-scale stock market simulation with generative agents, appears at ICML 2026.
We introduce AgentSociety 2, an integrated research environment for executable social science.
Select both to include either role.
No matching publications. Try another keyword, year, or author role.
Department of Electronic Engineering · Advisor: Prof. Yong Li
Information and Communication Engineering · Advisor: Prof. Yong Li
Electronics and Communication Engineering · Advisor: Prof. Yong Li
Communication Engineering
I welcome conversations and collaborations on social intelligence, human–AI systems, and computational experiments with artificial societies.