From Framework to Nodes: Investigating Semantic Interaction in Generative AI Dialogue Flows
Yanyang Ma, Kerun Li, Yulei Liu · International Journal of Human-Computer Interaction · 2026
This study adopts a modular perspective. It deconstructs generative AI dialogue flows into two dimensions: an interaction framework and a semantic framework. The research investigates user experience at three core interaction nodes—Confirmation, Warning, and Error. We integrated the Theory of Planned Behavior, Importance Theory, and Risk Perception Theory. Using these, we constructed a user experience evaluation model for dialogue nodes. First, we collected user perception data through experiments involving 12 typical interaction scenarios. We then verified the data clustering. Subsequently, we applied Structural Equation Modeling (SEM) and Bootstrap sampling tests. Finally, we discuss the interaction framework at various levels. Based on the findings, we propose design strategies for AI dialogue nodes. This research not only clarifies the user acceptance mechanisms for three types of special dialogue nodes but also provides an interaction framework reference and theoretical support for the design and evaluation of intelligent dialogue systems.