From Experience to Adoption: Modelling EFL Learners' GenAI Continuance Through Linear and Configurational Approaches

ABSTRACT As generative artificial intelligence (GenAI) increasingly penetrates language education, understanding learners' continued intention to use this technology becomes crucial. This study examines EFL learners' continuance intention to use GenAI for language learning through PLS‐SEM and fsQCA methodologies. Participants were undergraduate EFL learners ( N = 383) from three universities in Taiwan, aged 18–24, with varying English proficiency levels (TOEIC 450–860 or CEFR A2‐B2). Data were collected through an online survey comprising validated scales measuring technology acceptance constructs and personality traits using 5‐point Likert scales. Analysis employed both partial least squares structural equation modelling (PLS‐SEM) and fuzzy‐set qualitative comparative analysis (fsQCA). Results showed that three key dimensions – dependability, attractiveness and novelty – shape users' overall experience with GenAI, where dependability and attractiveness specifically align with users' initial expectations. Both expectancy confirmation and social influence predict learners' continuance intention. Openness to experience moderates these relationships positively for expectancy confirmation but negatively for social influence, indicating that individuals with high openness prioritise personal evaluation over social cues. As the first study to combine PLS‐SEM with fsQCA in GenAI use research, it reveals complementary insights: while PLS‐SEM identifies key predictors of continued use, fsQCA uncovers multiple pathways to adoption, with dependability and social influence as core conditions across configurations. This dual‐method approach advances theoretical understanding of technology adoption by demonstrating how linear relationships coexist with complex configurational patterns. The findings provide practical implications for educational technology design, emphasising the need to balance system reliability and aesthetic appeal while accounting for individual differences in user personality traits.