Research (Submitted / Ready for Submission)


Multi-Agent Reinforcement Learning in the Labor Market (under review; title and abstract anonymized for peer review) with Yancheng Liang, Boyang Zhou, Jiayi Yuan, and Natasha Jaques. First & Corresponding Author.

A simplified anonymous summary: We study two-sided matching markets under imperfect information and dynamic interaction. While classical stable matching models assume fully observable and fixed match quality, many real-world markets involve costly information acquisition, uncertainty, and strategic adaptation over time. Our work explores how multi-agent reinforcement learning can serve as a simulation framework for analyzing matching markets with endogenous information frictions. We develop a configurable environment. The results suggest that when information must be learned over time, decentralized learning dynamics can generate outcomes that differ meaningfully from traditional static matching mechanisms. More broadly, this work provides a computational approach to studying strategic behavior and equilibrium formation in dynamic two-sided markets.


Between Bytes and Bonds: A Bond Market Perspective on the Social Impact of Cryptocurrencies with Yanzhen Chen and Shu He.

Abstract: Cryptocurrencies, recognized for their inherent anonymity and affordability, present a conundrum with potential societal impacts. While they face scrutiny due to their association with illicit activities, their potential to empower marginalized demographics and invigorate local economies through innovation cannot be overlooked. This dichotomy necessitates an empirical evaluation of the true socioeconomic repercussions of cryptocurrency adoption. In this study, we employ staggered Difference-in-Difference models and analyze the cost of borrowing for municipal bonds with the entry of 1,800 cryptocurrency ATMs up to December 2019 to interpret the overarching influence of such technological adoption on local economies. Moreover, we delineate various impact channels. One notable adverse effect is the increased local crime and drug addiction. In contrast, positive outcomes manifest in the form of heightened innovation and the financial enfranchisement of marginalized groups. This research underscores the paramount need for timely insights into the multifaceted impacts of cryptocurrencies, aiding policymakers and regulators in formulating informed strategies in this rapidly evolving digital ecosystem.