Learning & Decision-Making in Societal Systems: Theory, Algorithms, and Design
Speaker: Eric Mazumdar, University of California, Berkeley
Date: March 31, 2021, noon
The ability to learn from data and make decisions in real-time has led to the rapid deployment of machine learning algorithms across many aspects of everyday life. Despite their potential to enable new services and address persistent societal issues, the widespread use of these algorithms has led to unintended consequences like flash crashes in financial markets or price collusion on e-commerce platforms. These consequences are the inevitable result of deploying algorithms--- that were designed to operate in isolation--- in uncertain dynamic environments in which they interact with other autonomous agents, algorithms, and human decision makers.
To address these issues, it is necessary to develop an understanding of the fundamental limits of learning algorithms in societal-scale systems. In this talk, I will give an overview of my work on three aspects of learning and decision-making in societal-scale systems: (i) Model-based decision-making in uncertain dynamic environments, (ii) Learning expressive models of human decision-making from data, and (iii) Understanding why and when current machine learning algorithms fail in game theoretic settings.
Eric Mazumdar is a Ph.D candidate at UC Berkeley advised by Michael Jordan and Shankar Sastry. His research lies broadly at the intersection of statistical machine learning, stochastic control, and economics with a particular focus on understanding the fundamental limits of learning algorithms in societal-scale (multi-agent) systems. Prior to his Ph.D he received an S.B in electrical engineering and computer science from MIT.