Dynamic Pricing by Software Agents

Monday, December 4, 2000 - 11:00am - 12:00pm
Keller 3-180
Jeffrey Kephart (IBM)
We envision a future in which the world economy and the Internet will merge and evolve into an information economy bustling with billions of economically motivated software agents that exchange information goods and services with humans and other agents. These software agents will constitute a new economic species that differs in some significant ways from humans. What impact might collective interactions among agents have on market behavior, and on the global economy as a whole?

In this talk, I will focus on collective interactions among software agents that dynamically price information goods or services. I will discuss two types of dynamic pricing. First, I will cover a few examples of dynamic posted pricing, in which automated pricing agents (pricebots) continually adjust the price of a commodity (or the price and attributes of a more complex product) in an effort to maximize profits. The price dynamics that result from interactions among pricebots are often quite complex, sometimes to the detriment of buyers as well as sellers. I will describe the connection of these and other nonlinear dynamical effects to optimization and learning. Second, I will discuss some very recent work on bidding algorithms for agents. We have found that relatively simple algorithms for a multi-unit continuous double auction consistently outperform humans in controlled experiments involving students and IBM Researchers.