Learning in GenAI ‐Mediated Education: A Hybrid SEM – PNA Approach to English Majors' Smart Service Interactional Experience and Anxiety

ABSTRACT Generative artificial intelligence (GenAI) has reshaped EFL learning, yet its extensive use has also produced distinctive anxieties. Few studies have examined how such anxiety emerges from the holistic interaction between learners and intelligent systems and fewer still have drawn pedagogical implications. Guided by Hassenzahl's Model of Pragmatic and Hedonic Quality of User Experience, we investigated how eight smart service interactional experience (SSIE) dimensions related to GenAI‐mediated EFL anxiety among 478 English majors across three Chinese universities. Using SEM (AMOS 24) and psychological network analysis (PNA, R 4.5.3), it was found that Ease to Use and Flow significantly and negatively predicted anxiety, whereas Efficiency and Newness unexpectedly exerted positive effects. In the network, only Flow maintained a stable negative association with anxiety and emerged as the primary bridging node linking SSIE dimensions to anxiety, while Ease to Use functioned as the global hub of the overall SSIE network. These findings suggest that the anxiety‐inducing effects of Efficiency and Newness may operate via disrupting Flow. Pedagogically, cultivating English majors' deep engagement in GenAI‐assisted tasks is more critical for anxiety reduction than optimizing efficiency or novelty, with instructional design favouring uninterrupted, learner‐paced interaction and metacognitive training to decouple self‐evaluation from system output.