认知心理学启发的多角色协同群智说服对话  

Multi-agent collaborative crowd intelligence persuasive dialogue inspired by cognitive psychology

作　　者：陈梦琦 郭斌[1] 王豪 刘静琦 赵凯星 梁韵基[1] 潘岩 於志文[1] Mengqi CHEN;Bin GUO;Hao WANG;Jingqi LIU;Kaixing ZHAO;Yunji LIANG;Yan PAN;Zhiwen YU(School of Computer Science,Northwestern Polytechnical University,Xi'an 710129,China;College of Systems Engineering,National University of Defense Technology,Changsha 410073,China)

机构地区：[1]西北工业大学计算机学院,西安710129 [2]国防科技大学系统工程学院,长沙410073

出　　处：《中国科学:信息科学》2025年第4期860-886,共27页Scientia Sinica(Informationis)

基　　金：国家杰出青年科学基金(批准号:62025205);国家自然科学基金(批准号:62032020)资助项目。

摘　　要：说服是人类交互过程中的重要能力,赋予人机对话交互系统强大的说服能力,可以实现对交互过程中话题走向和主观逻辑的有效控制,促进人机交互系统进一步智能化.现有说服性人机对话交互研究缺乏对说服机理层面的建模,且仅关注于一对一的简单说服场景,难以适应复杂且更具有普适性的群体说服场景.鉴于此,本文提出“群智说服对话”的概念,利用多个角色各异的说服智能体,从不同的话题角度,结合多样化的群体说服策略,协同配合以高效完成说服任务.首先,本文从认知心理学基础理论出发,总结归纳可以有效改变说服者心理认知的群智说服对话策略.其次,对复杂群智交互场景下的说服过程进行深度剖析,建立面向复杂说服场景的通用群智说服对话系统架构,实现对群智说服对话场景的深入理解和全面建模.然后,本文对群智说服对话领域的研究挑战和关键技术进行了阐述和讨论.最后,介绍本文在这方面的研究实践.Persuasion is a pivotal skill in human interaction,and endowing human-machine dialogue systems with robust persuasive capabilities can effectively control the direction and subjective logic of interactions,thus fostering further intelligence in human-machine interactions.Current research in persuasive human-machine dialogue lacks a model of the persuasive mechanism itself and focuses only on simple one-on-one persuasive scenarios,which do not adapt well to complex and more universally applicable crowd persuasion contexts.In light of this,this paper introduces the concept of“crowd intelligence persuasion dialogue,”employing a variety of persuasive agents with distinct roles.These agents collaboratively employ diversified crowd persuasion strategies from different topical perspectives to efficiently accomplish persuasion tasks.Initially,this study begins with foundational theories in cognitive psychology to summarize and deduce dialogue strategies of crowd intelligence persuasion that can effectively alter the persuader’s cognitive perceptions.Following this,a deep analysis of the persuasive process in complex crowd interaction scenarios is conducted,establishing a general system architecture for crowd intelligence persuasion dialogue tailored to intricate scenarios.This facilitates a comprehensive understanding and modeling of crowd intelligence persuasion dialogue scenarios.Subsequently,this paper discusses the research challenges and key technologies in the field of crowd intelligence persuasion dialogue and concludes with an overview of our research practices in this area.

关 键 词：群智说服对话 说服策略 群智能体 认知心理学 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]