Key points are not available for this paper at this time.
While generative artificial intelligence (GAI) holds promise for supporting academic language and content learning, its role in CLIL science instruction remains underexplored. This mixed-methods study investigates how two GAI-supported approaches, Inquiry-Based Learning (IBL) and Metacognitive Scaffolding (MS), influence fifth-grade students’ vocabulary acquisition, content comprehension, lab report writing, and learning perceptions. Sixty-six students from three Taiwanese elementary schools participated in a 10-week intervention using ChatGPT 4.0 within either IBL (n = 34) or MS (n = 32) frameworks. Data included pre- and post-tests, lab reports, and bilingual survey responses. Both groups made significant gains in vocabulary and content knowledge. However, IBL students demonstrated stronger vocabulary growth and deeper conceptual integration, reflected in more precise terminology and analytical writing. MS students showed clearer procedural structure but tended toward surface-level analysis. While students in both groups viewed ChatGPT positively, they also reported occasional difficulty with language complexity. This study contributes empirical evidence and theory-informed design insights for integrating GAI into CLIL instruction. It demonstrates how aligning AI prompts with constructivist (IBL) or metacognitive (MS) principles can promote vocabulary use, scientific reasoning, and reflective writing. Findings support a scalable framework that combines inquiry-driven and metacognitive scaffolds to enhance language-content integration in bilingual science classrooms.
Cheng-Ji Lai (Tue,) studied this question.