[3O2-IS-3] Agents

12:00 PM - 12:15 PM JST(3:00 AM - 3:15 AM UTC)

[3O2-IS-3-05] Personalized Persuasion in Human-AI Dialogue: A Dual-Route Model Integrating Cognition, Emotion, and Value Alignment

〇Siqi Lyu¹, Kazunori Terada¹ (1. Gifu University)

work-in-progress

Keywords:

Multi-Attribute Utility Model,Elaboration Likelihood Model,Persuasive strategy

Persuasive effectiveness depends on how rational cognition and affective motivation are balanced in communication. We introduce a computational framework that models this balance by integrating the Multi-Attribute Utility Model (MAU) with the Elaboration Likelihood Model (ELM) in a large language model (LLM)-driven agent. The system personalizes messages according to each individual’s belief-value structure and adaptively coordinates central and peripheral processing routes. In an experiment with 140 participants discussing fully autonomous vehicles, three strategies (Positive, Neutral, Negative) were tested. The Positive strategy enhanced both rational beliefs and experiential desires, while the Neutral condition improved only beliefs and the Negative condition showed no effect. These results reveal how psychological distance and value alignment shape persuasion in human-AI dialogue, providing a unified account of cognitive and affective mechanisms in adaptive communication.