A Human-Computer Negotiation Model Based on Q-Learning

The proliferation of artificial intelligence in the e-commerce industry has created new opportunities for online transaction negotiation. Traditional online dialogue systems are unsuited for this task, as they cannot negotiate and analyse user behaviour. To address the issue, this paper proposes a human-computer negotiation model that integrates the architecture of a dialogue system with automatic negotiation. The system is developed based on Q-learning and deployed on the WeChat applet platform. It can handle repetitive and diversified negotiations, negotiate with multiple parties, improve negotiation efficiency and reduce manual customer service costs. We have done extensive experiments to show that the system can negotiate efficiently with human users.