Building intelligent human-machine dialogue agents that can conduct natural and engaging conversations with humans is the long-standing goal of artificial intelligence (AI). Moreover, the persuasive ability of dialogue agents has garnered extensive attention from researchers. Persuasion is one of the crucial abilities in human communication. The Elaboration Likelihood Model (ELM) theory suggests that people tend to engage with persuasive messages when communicating with others. It is a prevalent phenomenon for individuals to hold diverse perspectives on a given topic and endeavor to influence others in altering their viewpoints, attitudes, or behaviors through conversational interactions. The development of intelligent persuasive dialogue agents that can persuade users to accept certain standpoints is emerging as a promising research field. There have been significant advancements in persuasive dialogue systems, which primarily enhance persuasiveness from three perspectives: integrating persuasion strategies, planning topic paths, and extracting argument structures. In this paper, we argue that persuasion is a cognitive psychology activity and that the persuasion strategy, the topic path planning strategy, and the argument structure prediction strategy can all be categorized as cognitive strategies. We define cognitive strategy-enhanced persuasive dialogue agent as CogAgent. As an emerging research area, an in-depth survey of the existing academic efforts is necessary.
To address these issues, a research team led by Bin Guo published their new findings on 15 May 2025 in Frontiers of Computer Science, a journal co-published by Higher Education Press and Springer Nature.
The team conducted a comprehensive review of concepts, challenges, methodologies, and applications in the field of cognitive strategy-enhanced persuasive dialogue. They formalize the definition of cognitive strategies extended from cognitive psychology theory. Based on the formalized concept model and generic system architecture, they summarize representative research in CogAgent from a systematic perspective. Furthermore, benchmarks, evaluation metrics, and thoughts on promising research trends are analyzed to promote the research progress.
DOI: 10.1007/s11704-024-40057-x