Impacts of AI-enhanced task-based learning on EFL postgraduates’ higher order thinking skills and English academic writing self-efficacy

Abstract

This study investigates how AI-enhanced task-based learning (AI-TBL) affects postgraduate EFL students' higher-order thinking skills (HOTS) and academic writing self-efficacy (SEF) compared with traditional task-based learning (TBL). Guided by constructivist learning theory, a 14-week quasi-experimental design was implemented with 168 postgraduate participants from various disciplines, all enrolled in an English for Academic Purposes (EAP) course at a Chinese university. The experimental group engaged in AI-enhanced instruction via the Yuketang platform, integrating AI-adaptive prompting, AI-generative feedback and AI-supported collaboration, while the control group followed conventional TBL without AI. Data were analyzed using MANCOVA, ANCOVA, and Epistemic Network Analysis (ENA). Findings revealed that AI-TBL significantly improved students’ critical thinking and problem-solving abilities, though its impact on creativity was limited. In academic writing self-efficacy, notable gains emerged in linguistic knowledge, information organization, writing performance, and self-regulation, whereas rehearsal and memory efficacy remained unchanged. The Epistemic Network Analysis (ENA) results reveal that AI-TBL fostered denser interconnections between HOTS and SEF, forming a cohesive structure linking critical thinking, problem-solving, linguistic confidence, and metacognitive control. However, the impact on creativity and rehearsal-memory efficacy was minimal. The study contributes to understanding how AI scaffolding facilitates the relationship between HOTS and SEF in task-based pedagogy, supporting deeper engagement and metacognitive control. Implications for AI-enhanced EAP instruction emphasize the balance between technological adaptivity and human-guided creativity and reflection.