Zachary Kessler

Postdoctoral Research Associate

Curriculum vitae

Policy Modeling, Public Policy Programme

Alan Turing Institute


Using AI in Economics and Economics to Understand AI

My research is generally dedicated to two main threads. The first utilizes modern computational and data advances to understand economic systems and model their dynamics. Using AI in a variety of forms, such as reinforcement learning, machine learning, and agent-based models, and coupled with more traditional techniques, I seek to understand the economy and potentially improve economic policy and institutions. The second examines what potential insights economics can offer to artificial intelligence. Here I apply tools from economic fields like political economy, institutional analysis, and labor economics to understand both the expected impact of greater prevalence of AI as well as methods to improve it.

In the first thread, my work has two main components at present. First, I am building simulation models to better understand the nature of labor sorting in environments with high degrees of complementarity and account for previously unknown, potentially inefficient outcomes a market can generate in such contexts. Second, I also program human institutions as computational algorithms to understand their role in generating observed equilibria and steady-states. From this position, using tools from computer science, I design experiments to test the efficiency and effectiveness of this human "algorithm". I apply this approach to estimate empirical models, such as in my work on HOA governance, as well as provide clear statements on the economic effect of transitioning between institution types, such as my approach to analyzing central banks.

In the second thread, I currently am using tools from political economy and public choice economics to translate human mechanisms meant to align goals of various agents in political settings to design algorithms to assist in resolving challenges  with instrumental convergence, power-seeking, and general alignment pursuits. Additionally, I develop a framework for predicting potential areas where AI and human labor possess comparative advantages and outline the implications for the future of work.