Many auctions used in practice are extremely simple, but do not satisfy the usual standards of mechanism design. These auctions are referred to as "auction games. " The Internet provides an environment running millions of auctions, an environment where simplicity is more important than ever before. The "ideal" of mechanism design is the single item Vickrey auction (second price auction), selling a single item to the highest bidder at the second highest price. This auction has all the desired features combining simplicity with efficiency: the auction has simple and intuitive rules; it has simple bidding strategies, as truthfully reporting the agent's value as the bid is dominant strategy; and it leads to efficient allocation. A main theme in mechanism design has been designing such truthful and efficient auctions in settings other than the single item auction. Unfortunately, truthfulness and efficiency often come at a price: the resulting mechanisms can be too complex for many environments.
The goal of this project is to develop tools to understand outcomes of auction games, such as Ad-Auctions (a multi-billion-dollar game played by Internet service providers and advertisers), and to quantify the expected efficiency and revenue. The Principal Investigator will investigate this goal under various models of the information structure of the game, as well as under various assumptions on the rationality of the agents.
The broader impact of the proposal is in developing the basic tools needed to understand many natural auctions, such as the auctions typically being used in practice. Such understanding will lead to simple and intuitive auctions in different environments that result in good quality outcomes.
