Testing a Purportedly More Learnable Auction Mechanism
Executive Summary — Each year, auctions are used to determine how billions of dollars of goods and services will be allocated across the globe. On eBay alone, $52.5 billion in merchandise was exchanged in 2.4 billion auctions conducted during fiscal year 2006. Considerable attention has been paid in the academic literature to the question of how to design auctions with efficient allocation and revenue-maximizing properties. However, in part because auction rules are typically published and standard theory assumes economic agents are capable of computing optimal strategies from published rules, little attention has been paid to the question of how to design auctions whose optimal strategies are easy to learn. Evidence suggests that even when auction rules are published and dominant strategies exist, people nonetheless struggle and sometimes fail to learn to play their optimal strategy. As a result, the authors argue that the question of how to design a learnable, strategy-proof auction mechanism is an important one. Key concepts include:
- Designers of auction mechanisms should create mechanisms that are easier for people to learn.
- This paper describes an auction mechanism that has received attention in the computer science literature because of its theoretical property of being more learnable than the standard mechanism. In fact, the new mechanism produced slower learning in human subjects than the standard mechanism.
We describe an auction mechanism in the class of Groves mechanisms that has received attention in the computer science literature because of its theoretical property of being more "learnable" than the standard second price auction mechanism. We bring this mechanism, which we refer to as the "clamped second price auction mechanism," into the laboratory to determine whether it helps human subjects learn to play their optimal strategy faster than the standard second price auction mechanism. Contrary to earlier results within computer science using simulated reinforcement learning agents, we find that both in settings where subjects are given complete information about auction payoff rules and in settings where they are given no information about auction payoff rules, subjects converge on playing their optimal strategy significantly faster in sequential auctions conducted with a standard second price auction mechanism than with a clamped second price auction mechanism. We conclude that while it is important for mechanism designers to think more about creating learnable mechanisms, the clamped second price auction mechanism in fact produces slower learning in human subjects than the standard second price auction mechanism. Our results also serve to highlight differences in behavior between simulated agents and human bidders that mechanism designers should take into account before placing too much faith in simulations to test the performance of mechanisms intended for human use.