Project Details

Description

Increasingly, automated economic transaction systems in eBay, Google and other institutions negotiate, buy and sell goods, services, advertisements etc. They use auctions to make decisions including pricing, allocation and optimization. As a result, we now have auctions systems far greater in scale than the traditional 'human scale' of negotiations and specialized auctions, and they impact our lives in sophisticated ways. These systems face many algorithmic challenges in the interface of Economics, Learning Theory, and Optimization, which is the focus of this project.In this project researchers will (a) design and analyze models for the various parties (users, auctioneer, buyer and seller) and their impact on the auctions; (b) design and analyze mechanisms in the presence of parties with mixed utilities that go beyond the traditional linear profit; (c) quantify impacts of budgets in mechanisms on truthfulness, equilibria and utilities, which has been traditionally underemphasized; (d) study the effect of bounded computational power and rationality on mechanisms; (e) design richer mechanisms for futures, combinatorial goods as well as dynamic settings; (f) study privacy, security and verifiability of auction mechanisms; (g) study the various learning and optimization problems that are fundamental to the tasks above.This project ultimately addresses the questions of how various parties with natural knobs (budget, utility) interact with automated economic transaction systems, how information is learned, used and controlled in such systems, and how these systems will evolve over the long term. The project explores these questions via the specific research tasks above, as well as via training undergraduate and graduate students to work in the interface of Economics, Optimization and other areas.
StatusFinished
Effective start/end date9/1/098/31/12

Funding

  • National Science Foundation (National Science Foundation (NSF))

Fingerprint Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.