Intelligent Agents for Multi-Issue Auctions and Bidding

An approach to auctions and bidding is founded on observations and expectations of the opponents' behavior and not on assumptions concerning the opponents' motivations or internal reasoning. The approach draws ideas from information theory. A bidding agent employs maximum entropy inference to determine its actions on the basis of this uncertain data. Maximum entropy inference may be applied both to multi-issue and to single-issue negotiation. Multi-issue variants of the four common auction mechanisms are discussed.