Emotion-driven Reasoning Model for Agent-Based Human–computer Negotiation

Jinghua Wu,Ting Zhang,Yan Li,Guangjuan Zhou

DOI: https://doi.org/10.1016/j.eswa.2023.122448

IF: 8.5

2024-01-01

Expert Systems with Applications

Abstract:Human-computer negotiation plays an important role in today's highly developed online transactions. Agents need to have higher emotional intelligence to deal with the random actions of human adversaries. Most of the existing studies focus on deterministic modeling but neglect the role of fuzziness in human-computer negotiation and emotion effects on the interaction between persuasion strategy selection and issue updating. This paper uses emotional rendering to describe the emotional interaction between agents to supplement the emotion generation process in agent-based negotiation. Moreover, this paper constructs emotion-driven reasoning model which integrates an agent’s assessment of emotion and actual utility, and proposes a mechanism for the selection of persuasion strategies and negotiating issue updating driven by the agent’s reasoning ability. A series of agent-agent experiments are conducted, and a human-computer negotiation system is developed for the real human-agent experiment. The agent-agent experimental results show that, compared with the non-emotion model, the proposed model increases the joint utility by 55.1% and compared with the non-reasoning model, the proposed model improves the success rate by 49.2%. Meanwhile, the proposed model is superior to several existing competing models in terms of the success rate, the negotiating rounds, and the success utility. The human-agent experimental results show that the developed negotiation system can achieve satisfactory results with human opponents.