The effect of online opponent modeling on utilities of agents in bilateral negotiation

Abstract

Negotiation is a communication process in which different parties try to reach a common agreement. Due to high cost and time spent on traditional negotiation, in the last two decades automated negotiation has been considered. Similarly, in an automated negotiation, competing parties often do not reveal their complete or true preferences. Such setting is called an incomplete information environment. To overcome the complexity that it generates, agents can try to use online opponent modeling, learning the preferences of the opponent during the negotiation. This paper tries to find settings in which the opponent modeling helps agents to improve their performance in a bilateral negotiation. The results of the experiments show that the use of modeling by one or both of the agents will definitely improve social welfare. But when one agent uses opponent modeling, its utility is not necessarily increased.