Michael Albert

Data driven mechanism design

My research focuses on the intersection of machine learning, data driven optimization, and mechanism design for systems of interacting, self-interested, strategic agents.

I am the John Tyler Associate Professor of Business Administration at the Darden School of Business at the University of Virginia, where I have also held appointments in Computer Science and Engineering Systems and Environments in the School of Engineering and Applied Science. I am a Batten Faculty Fellow and a Fellow of the LaCross AI Institute's Fellowship in AI Research (FAIR) program. I'm motivated by practical problems at the intersection of learning, optimization, computation, and economics, specifically the design of markets (or mechanisms) and learning in the presence of strategic agents. As the world moves towards increasing automation, there are, more and more, opportunities to combine the wealth of data with a principled, provably optimal approach to mechanism design in order to make intractable problems tractable and impossible problems possible.

Research

For my near term research agenda, I am interested in application areas where new technologies are allowing for transformational opportunities for mechanisms. Most recently, the rise of large language models and generative AI has created entirely new markets, for example for the data and content that these systems depend on, and the question of how to design those markets is largely open. Similarly, the abundance of data and modern machine learning have made it possible to tightly couple prediction and optimization in operational decisions such as pricing, which raises new questions about how to make those decisions robust to errors in the underlying predictions. Earlier in my career, I worked on two other such areas: the introduction of self-driving, connected vehicles, which provides a once in a generation opportunity to revamp traffic infrastructure, and the move to a "smart" grid for the distribution of electricity, which allows for a degree of sensitivity to incentives that, up to now, has been impossible in energy markets. Please look at my publications for more information.

Longer term, I am very interested in machine learning questions in settings where the agents generating the data have incentives to strategically manipulate the data. This line of research requires combining traditional algorithms in machine learning with techniques from mechanism design to provably generate accurate data and, more broadly, understanding when and how a decision maker can learn effectively through repeated interactions with strategic agents.

View Publications »

Contact Information

University of Virginia
Darden School of Business
100 Darden Boulevard
Charlottesville, VA 22903
Phone: (434) 243-2859
Email: albertm@darden.virginia.edu
Website: www.michaelalbert.co

Send Email »