Integrating Cognitive Processing, Emotional Engagement, and Belief–Value Alignment in Personalized Persuasive Human–AI Dialogue
Abstract
As an important capacity in human communication, the formation and effectiveness of persuasion depend on the dynamic coordination between affective motivation and rational cognition. Based on a questionnaire survey involving 140 participants, we introduced a persuasion framework to model this balancing relationship. Within this framework, the Multi-Attribute Utility Model (MAU) was used to represent the dimension of rational cognition, the Elaboration Likelihood Model (ELM) was used to represent the dimension of affective motivation, and both were integrated into the persuasive generation process of the large language model (LLM). In this study, we examined participants' attitudes toward fully autonomous vehicles (FAVs) and established three persuasive strategies (Positive strategy, Neutral strategy, Negative strategy) to investigate the effects of different strategies on rational beliefs and experiential desires. The results indicate that the Positive strategy not only significantly improved participants' rational beliefs, but also enhanced their experiential desires; the Neutral strategy primarily improved rational beliefs; whereas the Negative strategy did not produce a significant overall effect. These results reveal how psychological distance and value orientation influence persuasion in human–AI dialogue and provide a unified explanation for the cognitive and affective mechanisms in adaptive communication.