B2C电子商务人机谈判策略与系统研究

Human-Computer negotiation is an important branch of automated negotiation study, and is also an important method to implement the automatic trade in B2C e-commerce. However, few studies have shed light on the human-computer negotiation system. The core issue of automated negotiation study is the strategy designing, but the previous designed strategies lack the ability to cope with the random and changeable negotiation situation, such as the human’s negotiation behaviour, thus cannot meet the requirements of human-computer negotiation. Our study combines the time-dependent and behavior-dependent strategy to construct a combined strategy model, which is a critical method to build the human-computer negotiation system, and proposes a strategy selection algorithm, which can enable the agent to dynamically select the appropriate strategy according to the human’s negotiation behavior, rather than just keeping on one strategy from beginning to end, accordingly increasing the robustness and the flexibility of the negotiation system to a maximum extent. Based on the combined strategy selection model, a concession mechanism along the Pareto optimal front will be designed, facilitating the multi-attribute negotiation. The argumentation based negotiation technology will be integrated with the price negotiation, thus make the design and development of human-computer negotiation to be close to the practical application of the e-commerce. Finally, human-computer negotiation experiments will be conducted on a prototype system. The effect of the system will be evaluated from the perspective of economics and social psychology. The empirical study results can provide valuable theories for the future practical application of human-agent negotiation system. This study will have important theoretical value and practical significance to the research of e-commerce oriented automatic trading system.

人机谈判是自动谈判研究的重要分支，是实现B2C电子商务自动交易的重要手段。研究的核心问题是机器方（Agent）的谈判策略。目前策略研究缺乏对随机、多变谈判情境的处理能力，因此，无法应对人的谈判行为，不能满足人机谈判的需要。本课题从人的谈判行为研究入手，将经典策略模型的时间依赖和行为依赖特性相结合，设计组合策略模型及策略选择算法，使Agent不局限于某一种策略，而是可以根据对手的谈判行为，动态地选择合适的策略予以应对，极大地提高谈判系统的鲁棒性和灵活性。在此基础上，设计Pareto最优前沿上的组合策略让步机制，将策略选择推广到多属性谈判中去，并将价格谈判同基于论据的谈判相结合，使人机谈判的电子商务实际应用成为可能。最后，依托原型系统进行人机谈判实验，从经济学和社会心理学视角综合评估系统的应用效果，其实证研究结果可为未来基于谈判的电子商务自动交易提供理论依据，具有重要的理论价值和现实意义！

人机谈判是自动谈判研究的重要分支，是实现B2C电子商务自动交易的重要手段。研究的核心问题是机器方(Agent)的谈判策略。针对现有策略研究缺乏对随机、多变谈判情境的处理能力，无法应对人的谈判行为，不能满足人机谈判的需要，本课题使用人的谈判行为理论解释、设计机器谈判策略，将经典策略模型的时间依赖和行为依赖特性相结合，设计组合策略模型及策略选择算法，使Agent不局限于某一种策略，而是可以根据对手的谈判行为，动态地选择合适的策略予以应对，并能够处理人的非理性、不合作、欺骗等谈判行为，极大地提高了谈判系统的鲁棒性和灵活性。在此基础上，设计了Pareto最优前沿上的组合策略让步机制，将策略选择推广到双边多属性谈判中去，并将价格谈判同基于论据的谈判相结合形成人机自然语言交流，使人机谈判的电子商务实际应用成为可能。最后，依托原型系统进行人机谈判实验，从经济学和社会心理学视角综合评估了系统的应用效果，其实证研究结果可为未来基于谈判的电子商务自动交易提供理论依据，具有重要的理论价值和现实意义!