An Adaptive Bilateral Negotiation Model Based on Bayesian Learning

Endowing the negotiation agent with a learning ability such that a more beneficial agreement might be obtained is increasingly gaining attention in agent negotiation research community. In this paper, we propose a novel bilateral negotiation model based on Bayesian learning to enable self-interested agents to adapt negotiation strategies dynamically during the negotiation process. Specifically, we assume that two agents negotiate over a single issue based on time-dependent tactic. The learning agent has a belief about the probability distribution of its opponent’s negotiation parameters (i.e., the deadline and reservation offer). By observing opponent’s historical offers and comparing them with the fitted offers derived from a regression analysis, the agent can revise its belief using the Bayesian updating rule and can correspondingly adapt its concession strategy to benefit itself. By being evaluated empirically, this model shows its effectiveness for the agent to learn the possible range of its opponent’s private information and alter its concession strategy adaptively, as a result a better negotiation outcome can be achieved.