The following are some
ideas for broad topics you might explore in your
project. Feel free to go in a completely different direction though --
creativity is encouraged, and I will take an expansive definition of
what kinds of things are relavent to this course.
The schedule for the project follows:
- March 12:
Email me a few paragraphs about an idea for your project,
who (if anyone) you will be working with. Only one email per group is
needed. I'll give you some feedback.
- April 1:
Turn in (by email) your mid-project report. This should be a ~ 5
page description of your project, including a survey of the related
work, how you hope to extend it, and a summary of what you have done so
- April 25: Project presentations in class.
- May 8:
Final written project is due. This should be a polished 10+ page report
written in the style of an academic paper, discussing and citing
related work, and explaining your results in detail.
can be done in groups of size 1 or 2. Only one copy of each deliverable
(proposal, mid-project report, final paper) is needed per group.
I recommend looking for
project ideas by investigating one of the following 3 literatures -- but proposing completely novel directions is encouraged:
(Approximate) Strategy-Proofness in Large Markets
The economics literature has considered the properties of "large
markets" for decades, with a recent resurgence of interest. However, in
this literature, the notion of a "large market" often requires much
more restrictive assumptions than the kinds of "large game" settings we
have been talking about: i.e. the type
may need to remain fixed while the number of
players tends to infinity, the number of players of each type must tend
to infinity, players may need to have their types drawn from a
(sometimes product) distribution, there may need to be smoothness
conditions on families of equilibria, etc. Can we take results from
this literature and strengthen them (by weakening the assumptions
needed, making them algorithmically constructive, making their solution
concepts more robust, etc.) using the tools of differential privacy?
A sampling of this literature follows (but this is not exhaustive) --
follow the references!
Incentives for Price Taking Behavior in Large Exchange Economies
-- Roberts and Postlewaite, Econometrica 1976.
- Open-Loop and Closed-Loop Equilibria in Dynamic Games with Many Players -- Fudenberg and Levine, Journal of Economic Theory 1988.
- Asymmetric Information Bargaining Problems with Many Agents -- Mailath and Postlewaite, Review of Economic Studies 1990.
to Efficiency in a Simple Market with Incomplete Information
-- Rustichini, Satterthwaite and Williams, Econometrica 1994.
- Approximately Competitive Equilibria in Large Finite Economies -- Jackson and Manelli, Journal of Economic Theory 1997.
- Informational Size and Incentive Compatibility -- McLean and Postlewaite, Econometrica 2002.
- Large Robust Games -- Kalai, Econometrica 2004.
- Marriage, Honesty, and Stability -- Immorlica and Mahdian, SODA 2005.
- Efficiency of Large Double Auctions -- Cripps and Swinkels, Econometrica 2006.
and Stability in Large Two-Sided Matching Markets -- Kojima
and Pathak, American Economic Review 2009.
- Incentives in the Probabilistic Serial Mechanism -- Kojima and Pathak, Journal of Economic Theory 2010.
- The Combinatorial Assignment Problem: Approximate Competitive Equilibrium from Equal Incomes -- Budish, Journal of Political Economy 2011.
- Strategy-Proofness in the Large -- Azevedo and Budish, 2013.
Economic Implications of Privacy
The economics literature has
recently become interested in the effects that different kinds of
information disclosure have on equilibrium outcomes in various
settings. This has resulted in a series of papers that study simple
qualitative kinds of privacy (i.e. providing perfect information vs. no
information about some purchase decision), and makes qualitative
conclusions about the effects of privacy on welfare and revenue. Can
you study some of the same kinds of questions while making the results
quantitative, by using differential privacy as a measure of privacy?
A sampling of this literature follows (but again -- not exhaustive):
Extending the Game Theory/Differential Privacy Literature
the papers linked from the main course website, which represent most of
the published work at the border of differential privacy and game
theory. Can you extend them in new directions? Are there other problems
studied in the purely game theoretic literature (see either economics
journals at large, or the proceedings of computer science conferences
like EC, STOC, FOCS, ITCS, and SODA) that might better be modeled with
agents having preferences for privacy?