Artificial Intelligence and Auction Design

What is it about?

Motivated by online advertising, we study the design of auctions played by algorithmic agents. We find that the first-price auction is prone to collusion, while the second-price auction is fully competitive. Through multiple experiments, we identify the force that critically separates the two auction formats. We show that providing bidders with additional information, namely the minimum-bid-to-win, after the auction restores competition even in the first-price regime.

Featured Image