Simulation of Automated Negotiation

Automatednegotiationis arguedto improvenegotiationoutcomes and reduce tra- action costs in electronic business and to enable sophisticated cooperation and - ordination between software agents in autonomous systems. However, due to the youth of the ?eld operative systems do not yet exist and scholars rely on simulation studies to evaluate potential system con?gurations for automated negotiation. We review the state of the art literature on simulation of automated negotiationalong its main components: the negotiation problem, the interaction protocol, and the so- ware agents. De?ciencies of existing approaches concerning the practical appli- tion in an open environment as the Internet, where automated negotiation proceeds fast, with different opponents, and for various negotiation problems, are identi?ed. To address these de?ciencies we model and simulate automated negotiation s- tems consisting of software agents and interaction protocols, which were proposed in negotiation and game theoretic mechansim design literature but not used in s- ulations yet. The proposed software agents follow generic offer generation and c- cession strategies and are governed by interaction protocols that allow agents to interrupt their strategy to avoid exploitation and unfavorable agreements. The - put for these automated negotiation systems are preferences elicited from human in negotiation experiments to guarantee realistically complex and diverse negot- tion problems. Simulation outcomes are compared across the different systems and to the results of the negotiation experiments along various outcome dimensions – proportion of (Pareto-optimal) agreements, dyadic and individual performance, and fairness.