Scientific approaches and techniques for negotiation : a game theoretic and artificial intelligence perspective

Due to the rapid growth of electronic environments (such as the Internet) much research is currently being performed on autonomous trading mechanisms. This report contains an overview of the current literature on negotiations in the fields of game theory and artificial intelligence (AI). Game theorists have successfully developed and analyzed a variety of bargaining models in the past decades. We give an extensive overview of this theoretical work. In particular, research performed in the fields of cooperative and non-cooperative bargaining, bargaining with incomplete information, and bargaining over multiple issues is evaluated. The use and shortcomings of game-theoretical concepts in practical applications is discussed. Simplifying assumptions frequently made in game-theoretical analyses, such as assumptions of perfect rationality and common knowledge, do not need to be made if the behavior of boundedly-rational negotiating agents is modeled directly, for instance using techniques borrowed from the field of AI. We show how different AI-techniques, such as decision trees, Q-learning, evolutionary algorithms, and Bayesian beliefs, can be used to develop a negotiation environment consisting of intelligent agents. These agents are able to adapt their negotiation strategies to changing user preferences and opponents. A survey of state-of-the art applications using AI-techniques is given in this report. The main conclusion from this survey is that combining techniques and ideas from game theory and AI will make it possible to create robust and intelligent negotiation systems in the near future.