Study on an Argumentation-Based Negotiation in Human-Computer Negotiation Service

With the rapid development of e-commerce, the automatic transaction has been a potential demand of enterprises. Concentrating on B2C e-commerce scenario, this paper designs a human-computer negotiation model and algorithm, through which an agent can negotiate with a human using natural language. To validate the model and algorithm, we conducted a between-group experiment based on a prototype system comparing the negotiation effect between the groups with and without arguments. The experimental results show that adding arguments into price negotiation can significantly increase the success rate of the human-computer negotiation. Moreover, there is a significant positive impact on the buyer’s subjective feelings on system using, as well as the seller agent’s economic utility, so that it can finally help both sides to reach a win-win situation in the negotiation. The contribution of our study can apply to B2C e-commerce platforms for improving the performance of their intelligent customer service agent. Our study is also meaningful for helping the two parties to increase their trading efficiency and decrease their trading cost.