Agent Negotiation under Uncertainty and Risk

Abstract

Traditional game theoretic reasoning for agent negotiation usually base on the assumption of rationality of agents who are expected utility maximizers. The utility functions that express preferences of agents over goods, states or money are essential in decision making of rational agents. However, the utility functions are very sensitive to agent’s wealth levels. To obtain the utility functions and wealth levels of other agents during the negotiation are extremely difficult. In this paper, we propose a way of getting around the problems by assuming the game theoretic decision making of rational agents be based on a monetary payoff game matrix instead of a utility payoff matrix. We regard utility functions and wealth levels of agents as private information while treating the monetary payoff game matrix as public information that is available to each agent. Rational agents of different risk preference types (e.g. risk averse, risk neutral and risk seeking) must negotiate to find a stable state using only the public information. We therefore extend the work of Wu and Soo who developed the negotiation mechanisms with a trusted third party as a mediator for agents to reach a stable equilibrium state under uncertain games. We discuss how the negotiation results based on the monetary payoff game matrix may be affected by different risk preferences of negotiating rational agents.