In one large, multi-team and ongoing project we design autonomous software agents that trade (on behalf of human designers) in complex electronic markets. New online markets (e.g., eBay, online brokerages), business procurement processes (e.g., reverse auctions), and dynamic trading of utility resources (e.g., electric power, pollution rights) is leading to unprecedented automation of markets, and with it opportunities for automated trading. Development of market procedures and institutions promoting efficient resource allocation or other social objectives is a central ICD problem, and this in turn requires accurate models of the trading strategies likely to be deployed in such markets. Analysis of trading games is complicated by enormous strategy spaces, incomplete information, and dynamic environments. We initiated an annual international Trading Agent Competition (TAC) to focus the research community on common scenarios, and to facilitate comparison and building on others’ ideas. Experience developing trading agents for the TAC travel-shopping gameand the supply chain management game has led to new insights and techniques, transferable to trading agent strategies for generic market mechanisms or other domains. Research on this challenging design task has also led to advances in strategic reasoning methodology, combining ideas from machine learning, simulation, and game-theoretic analysis.

More ICD in Action:

Trust, Reputation and Recommendation
Community Lab
Incentive-Centered Design for Spam
Market and Information System Design for Collaboration