Learner-AI interaction and prompting structures in argumentative continuation writing: a complex dynamic systems analysis of writing-thinking co-development

Guided by Complex Dynamic Systems Theory (CDST), this study examined how Chinese EFL learners’ interaction with generative AI (GenAI) was associated with the co-development of writing quality and critical thinking (CT) in argumentative continuation writing. Thirty-eight learners completed original writing and revised it with ChatGPT, and data included writing products, reflection journals, and interaction logs. Cluster analysis (CA) identified three developmental trajectories: (1) supportive development, with gains in writing quality and CT; (2) competitive development, with substantial CT gains but a slight decline in writing quality; and (3) restrained development, with declines in both domains. Epistemic Network Analysis (ENA) further modeled and visualized the distinct learner-AI interaction patterns underneath. Supportive development was associated with a context-oriented prompting structure, emphasizing its connection with instruction and polite language, competitive development with a strong instruction-output connection, and restrained development with a less coherent prompting structure. By making a CDST-informed use of CA and ENA, the study shows that writing-thinking co-development under GenAI mediation was contingent on how learners structured and regulated their interaction with AI. Theoretical, methodological, and pedagogical implications are discussed.