12 September 2016 2:25PM
The Departments of Economics and Computer Science have jointly funded a new postdoctoral position to support the growing economics and computation community at Duke University. Alumnus Michael Albert (Fuqua Ph.D. ’13) will be working primarily with Professors Charles Becker in the Department of Economics and Vincent Conitzer, who has appointments in the Departments of Computer Science and Economics.
“Much of economics is fundamentally concerned with the question of what influences the observed decisions that people make (i.e., what is a person’s utility function), and much of computer science — specifically artificial intelligence — is primarily concerned with how we can maximize the likelihood of achieving a specific goal or utility level. In many ways the two disciplines meet in the middle, and we can learn a lot by considering the act of decision-making from both sides,” Albert said. “I'm really excited about taking part in what is a unique cross-discipline research group working at this intersection. “
Albert also will be teaching a course for students in the M.S. Economics and Computation (MSEC) program related to his research area in the spring. As the MSEC program — now in its third year — continues to mature, the academic content is taking a more established form, with new courses and a regular CS-Econ Seminar Series. Many of the new and continuing students have been gravitating toward computational algorithms with economic applications, also known as mechanism design.
“This is the heart of the departments’ overlap — and, as in instructional program, I believe sets us apart from all other schools,” said Becker, who is one of MSEC’s co-director of graduate studies.
Albert is generally interested in designing mechanisms that are both sample efficient and computationally efficient to elicit information from users. More specifically, he wants to know whether computational economists can design better auctions based on how bidders value items. Albert is currently working on a research agenda related to the idea of correlated bidder valuations in auctions (for example, if one bidder has a high valuation for an item, the other is also likely to have a high valuation): a scenario in which mechanisms would enable the auctioneer to earn a lot more money than would be the case if the values were independent.
“We don’t see these mechanisms in practice because they require the auctioneer to know very precisely the distribution of bidders that are likely to show up. I’m working to design ‘robust’ mechanisms where there is a natural trade-off between how much the auctioneer knows and how much money he can generate from the auction,” he said.
Albert first developed an interest in “understanding decision-making in the context of systems” when he was an undergraduate at James Madison University (JMU). He graduated from JMU with bachelor’s degrees in quantitative finance and mathematics with a minor in physics in 2006. He further developed that interest during his time as a financial economics doctoral student at Duke’s Fuqua School of Business: “I started trying to understand the limitations of data in the mechanism design process; specifically, how to break one of the fundamental assumptions in much of the mechanism design literature — that of ‘common priors’ — without breaking it completely by incorporating data into the mechanism design process.” To complement this interest, Albert also graduated from Duke in 2013 with an M.S. in computer science.
His experience in economics and computer science, coupled with a strong quantitative background, will come in handy in his role as an instructor for MSEC students. For now, he advises: “Be eager to get involved! There a lot of good open questions, and bothering faculty — or postdocs — is a good way to figure out what some of those are!”
