AI-Driven Writing Instruction and College EFL Learners’ Writing Proficiency: A Complex Dynamic Systems Perspective

Generative artificial intelligence (GenAI) technology is transforming language education, particularly in the realm of writing instruction. Since the introduction of OpenAI’s ChatGPT, extensive research has demonstrated the efficacy of incorporating GenAI into writing instruction to boost learners’ writing proficiency. These studies have typically employed a “ladder” perspective to assess learners’ writing development through comparing the outcomes of pre- and post-intervention tests. However, this approach often overlooks the dynamic progression of writing competence. To address this gap, the present study investigated the developmental trajectory of tertiary-level English learners’ writing proficiency through the framework of Complex Dynamic Systems Theory (CDST). Over a 13-week AI-driven writing program, thirteen students participated and underwent seven writing assessments. The comparison between learners’ initial and final writing proficiency states revealed significant improvements in overall writing proficiency, as well as in the dimensions of writing complexity and accuracy, though not in fluency. Nonetheless, analysis of the writing outcomes indicated fluctuations in overall writing proficiency, complexity, fluency, and accuracy across the seven tests. Additionally, learners displayed individual variability in their developmental trajectories across all aspects of writing. The study also identified trade-offs among writing complexity, fluency, and accuracy throughout the instructional process. These findings provide empirical support for CDST within an AI-driven teaching context and offer valuable insights for enhancing writing instruction.