My research focuses on issues in the intersection of computer science
(especially artificial intelligence, multiagent systems, and computational
social choice) and economics (especially microeconomic theory, game theory,
and mechanism design). This includes the following topics.
(1) Making decisions based on the preferences of multiple agents.
This can be done using, for example, voting or auction mechanisms. I am
especially interested in cases where the space of alternatives is
combinatorial in nature, for example, when multiple items must be
allocated. This leads to questions about how the agents' preferences
should be represented and communicated, as well as nontrivial optimization
problems for finding the best alternative.
(2) Algorithms and complexity results for fundamental
game-theoretic solution concepts. Game theory studies how an agent should
act in the presence of other agents who have their own preferences. It has
introduced a variety of solution concepts, including Nash equilibrium but
also many others. I work on operationalizing these concepts, by designing
algorithms for computing or learning the solution in specific games. I am
particularly interested in settings where one party has a commitment
advantage, and these techniques are finding applications in real-world
(3) (Automated) mechanism design for strategic agents. In mechanisms such
as the ones discussed under (1), a selfish agent is sometimes better off
reporting her preferences inaccurately, in order to obtain an outcome that
is better for her. We can design these mechanisms in such a way that good
outcomes result in spite of selfish behavior. I am especially interested
in automated mechanism design (where we identify good mechanisms by having
a computer search through the space of possible mechanisms in an
intelligent way), and the effects of the agents' computational limitations
on the mechanism design process.
(4) Mechanism design in highly anonymous environments. Mechanism design
has traditionally focused on the problem of agents misreporting their
preferences. However, in highly anonymous environments such as the
Internet, another often-used manipulation is for an agent to pretend to be
multiple agents. I am working on extending the theory of mechanism design
to address this type of manipulation.