Trajectories of foreign language writing enjoyment in AI-mediated EFL contexts: A latent growth curve modeling study from a complex dynamic systems perspective

While AI tools are increasingly used in EFL writing instruction, longitudinal evidence on learners’ writing enjoyment remains limited. This study investigated the longitudinal development of foreign language writing enjoyment (FLWE) in an AI-mediated EFL writing context, drawing on Complex Dynamic Systems Theory (CDST). Adopting a mixed-methods longitudinal design, quantitative and qualitative data were collected from 320 undergraduate EFL learners over one semester. Quantitatively, latent growth curve modeling (LGCM) was conducted in AMOS 26 to examine overall developmental trajectories and individual differences in FLWE, as well as the predictive role of perceived classroom climate (PCC). Qualitatively, learners' reflective journals were analyzed using thematic analysis with independent double-coding to explore the mechanisms underlying different enjoyment trajectories and classroom-AI interaction experiences. The unconditional LGCM results indicated that FLWE remained relatively stable at the group level across the three measurement occasions, while significant individual differences were observed in both initial levels and rates of change. The conditional LGCM further showed that PCC significantly predicted learners' initial levels of FLWE and their developmental trajectories over time. Learners who perceived a more supportive and structured classroom climate reported higher initial enjoyment and greater growth in FLWE. Qualitative findings revealed heterogeneous emotional starting points and non-linear enjoyment trajectories, shaped by learners' evolving use of AI tools within a socially supportive classroom environment. Overall, the findings suggest that foreign language writing enjoyment in AI-mediated EFL contexts is a dynamic and individualized experience shaped by perceived classroom climate and classroom-AI interaction processes, and they also demonstrate the value of integrating latent growth curve modeling with reflective-journal thematic analysis to capture the complexity of affective development in technology-enhanced language learning.