The affective Human–AI Co-Flow model: an interaction design study of commuter behavior in AI navigation

Navigation service providers are shifting from functional tools toward intelligent agents with social roles and affective influence. Existing technology adoption research emphasizes cognitive and utilitarian dimensions, offering limited explanations of how AI develops from a tool into an affective partner during stressful commuting. This study proposes the Affective Human–AI Co-Flow (AHCF) model to explain how affective processes relate to navigation behavior through a sequence connecting AI Role Discourse, Affective Vulnerability Context, and Collaboration Fit with Affective Appraisal, Affective Regulation, and Affective Attachment. Using Amap’s “Gao Laoshi” AI agent as the empirical context, we collected data from 280 commuters and tested the model through structural equation modeling (SEM), artificial neural networks (ANN), and multi-group analysis (MGA). Collaboration Fit emerged as a key gateway in the affective sequence. After its inclusion, the direct effects of AI Role Discourse and Affective Vulnerability Context on Affective Appraisal became marginally significant or nonsignificant. Affective Appraisal was strongly coupled with Affective Regulation, which had significant positive structural relationships with both Behavioral Response and Affective Attachment. Differences in predictive rankings and group comparisons further indicated that AI’s affective influence varies with role perceptions, contextual pressure, and usage experience. AHCF thus reconceptualizes technology adoption as an affective co-flow process characterized by gating, amplification, and shifts across user groups. It provides a theoretical framework for understanding AI’s development from a functional navigation tool into a relational collaborator and offers emotion-centered implications for AI navigation and intelligent mobility.