Scaffolding L2 Learners’ Critical Thinking Tendency and Achievement Emotions: Students’ Experience of a Three-Stage AIassisted Speaking Pedagogy

While artificial intelligence (AI) tools offer rich input for language learning, few studies examine their role in driving higher-order cognitive processes, integrated pedagogical cycles, and broader achievement emotions. Addressing these gaps, this study investigated a scaffolded three-stage AIassisted speaking pedagogy (preparation, practice, and reflection) among 13 tertiary L2 learners (CEFR A2 speaking equivalent). Grounded in an integrated Community of Inquiry (CoI) and ControlValue Theory (CVT) framework, semi-structured interview data revealed that teaching, social, and cognitive presences form a dynamic, recursive system shaping students’ internal appraisals. Specifically, teaching presence established a supportive structural foundation, human-AI social presence functioned as an emotional engine buffering anxiety, and higher-order cognitive presence (integration) acted as a catalyst for high-value appraisals. Ultimately, this study proposes an integrated theoretical model showing how pedagogical scaffolding and dialogic AI interaction may collaboratively transform linguistic friction into empowering cognitive growth, offering practical insights for AI-driven language instruction.