Concession Strategy Adjustment in Automated Negotiation Problems

Yuchen Liu, Rafik Hadfi, Takayuki Ito

Abstract: Automated negotiation agents usually rely on theories and principles from other fields to guide their concession behavior so that they can perform better when put into productive environments. For example, a marketing agent developed for automated trading could rely on financial theories. While introducing new theories, however, new parameters will be introduced to the agent’s concession mechanisms as well. This paper, shows a method for adjusting these parameters to construct a more powerful concession mechanisms. Experiments were done with the Supply Chain Management League (SCML) one-shot environment, and the results indicate that this method can actually improve the performance of agents which employ theories mainly from economic fields. Furthermore, the method can also help distinguish models that are inefficient or even have negative effects in certain situations.