Peter R. Wurman
The majority of studies of agent decision making in environments with multiple auctions assume homogeneity in the rules of the auctions. However, online markets are much more fragmented, and products are often for sale in a variety of venues with different rules. We present a conceptual approach to designing e xible trading agents that is broad enough to encompass both multiple types of auctions and a variety of user preferences. The generalizations described lead to natural representations of the agent’s decision task as either a Markov decision processes or an extensive form game, depending upon the form of the market model.