Co-development of AI literacy, agency, enjoyment, and anxiety in EFL learners: A complex dynamic systems perspective

Artificial Intelligence (AI) tools have been found to influence English language learning. However, there is scant research on the role of AI literacy in learners' psycho-affective states. To address the gap, our study examines the longitudinal co-development of AI literacy, learner agency, foreign language enjoyment, and foreign language classroom anxiety among Chinese freshmen EFL learners over a 16-week semester. A total of 315 university students completed questionnaires on AI literacy, agency, enjoyment, and anxiety at weeks 1, 8 and 16. Multivariate latent growth curve modeling was used to estimate developmental trajectories and cross-construct coupling patterns. Results showed that AI literacy increased significantly over time, agency and enjoyment declined modestly, and anxiety decreased but not significantly. At baseline, AI literacy, agency, and enjoyment were positively interrelated, while anxiety showed negative associations with the other variables. In the later stage, growth trajectories of AI literacy, agency, and enjoyment were positively coupled, whereas anxiety exhibited weaker and partially independent dynamics. Significant negative intercept-slope associations indicated general catch-up patterns across constructs. Cross-domain associations indicated asymmetric relationships between initial levels and subsequent change in AI literacy and affective variables. Overall, the findings indicate that AI literacy may function as a dynamic cognitive resource associated with temporally uneven development in learner systems, with cognitive and affective subsystems evolving at different rates. Given the observational three-wave design, these patterns are interpreted as correlated rather than causal change. The study advances CDST-informed perspectives on AI-mediated language learning by highlighting the asynchronous nature of cognitive, emotional, and agentic development.