avatar

Haoyang Li

李昊阳

Incoming PhD Student
University of Virginia
lihaoyang@email.virginia.edu


About Me

I am an incoming PhD student in Systems Engineering at the University of Virginia (UVA). Currently, I am a final-year undergraduate at Hong Kong Baptist University (HKBU), majoring in Sociology with a minor in Computer Science. My studies have provided rigorous training in sociological theory and research methods, complemented by extensive coursework in computer science with a focus on Artificial Intelligence (AI).

Driven by a strong conviction in the transformative potential of interdisciplinary research, I am passionate about exploring the convergence of social science and AI in the era of big data and intelligent technologies. I am particularly interested in how these sociological theories-guided LLM agents can enable novel insights into social simulations and complex societal dynamics that both traditional data-driven machine learning and social science research cannot achieve.

I am currently supervised by Prof. Hongru Du at Department of Systems and Information Engineering, University of Virginia and collaborating with him to focus on the research of theory-guided LLM agent-based modeling and exploring its applications in domains including modeling human bahavior and public health. Priviously, I also serve as a Research Intern at Tsinghua University’s Future Intelligence Lab under the supervision of Prof. Fengli Xu and collaborated with Prof. James Evans at the Knowledge Lab, University of Chicago, where I focus on designing LLM-driven agents in physical environment and AI scientists in complex context.

For more details, please see my CV (last updated 01 Apr, 2026).

If you are interested in my work or have potential ideas to explore together, please feel free to email me. I would be happy to discuss further :-)

Research Interests

Research Experiences

News

Publications

  1. Theory-Informed Generative Agents for Human Mobility Modeling
    Haoyang Li, Runzhou Liu, Yao Li, Amy Wesolowski, Sen Pei, Hongru Du
    We introduce TIMA, a theory-informed generative agent framework that integrates physical mobility laws (EPR) with LLM semantic reasoning. It reconstructs city-scale mobility patterns and social segregation without training on ground-truth data, achieving a strong improvement in activity profiling accuracy over baselines. Furthermore, it enables counterfactual simulation, accurately reproducing the emergence of unequal home-stay behaviors and intensified stratification during the COVID-19 pandemic.
    PDF Code BibTex Under Review

  2. CrimeMind: Simulating Urban Crime with Multi-Modal LLM Agents
    Qingbin Zeng, Ruotong Zhao, Jinzhu Mao, Haoyang Li, Fengli Xu, Yong Li
    We proposed CrimeMind, a novel LLM-driven Agent-Based Modeling (ABM) framework for urban crime simulation. Integrating criminology and sociology theories into agent design, it improves crime hotspot prediction and spatial distribution accuracy by 24% versus traditional ABM and deep learning baselines.
    PDF Code BibTex Under Review

  3. The Emergence of Altruism in Large-Language-Model Agents Society
    Haoyang Li, Xiao Jia, Zhanzhan Zhao
    We introduce a Schelling-variant urban migration model to create a large-scale social dilemma for over 200 LLM agents, investigating the emergence of altruism. Our work provides the first evidence of a fundamental behavioral divide, identifying two archetypes: "Adaptive Egoists," who respond to social norms, and "Altruistic Optimizers," who inherently prioritize collective good. We show social interaction can increase altruistic actions significantly in egoistic models, proposing that the choice of LLM for social simulation is a choice of theoretical foundation.
    PDF BibTex Preprint

Awards & Honors