Affective pathways in human–AI interaction: the formation of willingness to communicate among Chinese KFL learners in AI-mediated informal digital learning of Korean

Abstract

This study examines willingness to communicate (WTC) among Chinese learners of Korean in the context of AI-mediated informal digital learning of Korean (AI-IDLK) from an affective perspective. Moving beyond the assumption that L2 confidence and anxiety function as parallel affective predictors, the study suggests that these variables are related to WTC in different ways in contexts of human–AI interaction. Using an explanatory sequential mixed-methods design, this study collected survey data from 205 Chinese KFL learners and analyzed them using SPSS 27.0 and the PROCESS Macro, including descriptive statistics, exploratory factor analysis, reliability testing, Pearson correlation analysis, hierarchical multiple regression, and mediation analysis. Interview data from nine participants were examined through NVivo-supported reflexive thematic analysis. The quantitative results show that AI-IDLK is positively associated with WTC, with confidence showing the only significant indirect association, while anxiety does not exhibit a significant indirect effect. Qualitative findings further help explain these results by showing that confidence is associated with Psychological safety, Repetitive production practice, Personalized adaptive learning, Sustainable accessibility, and Discourse agency and engagement, whereas anxiety persists in relation to Uncertainty of AI responses, Uncontrollability of interaction, Uncritical praise, and the Gap with real interaction. Taken together, these findings suggest that the affective foundation of WTC in AI-IDLK may not be fully explained by the conventional view of confidence and anxiety as parallel affective predictors, and they call for a refinement of WTC theory that more explicitly accounts for the affective conditions of human–AI interaction.