How Does Learner Prior Language Knowledge Play in GenAI-Assisted Project-Based Learning for Creative Thinking?

Abstract This study investigates how Generative AI (GenAI) scaffolding interacts with learners’ prior language knowledge to support creative thinking in project-based learning (PjBL). Grounded in Cognitive Load Theory and Aptitude-Treatment Interaction frameworks, a 10-week quasi-experimental study was conducted with 118 Chinese EFL learners (ages 13–15) who created English picture books using either supportive scaffolding (providing direct examples and structured guidance) or reflective scaffolding (offering open-ended prompts for metacognitive engagement). Moderated regression analyses revealed significant interaction effects: supportive scaffolding enhanced creative thinking and reduced cognitive load for lower-proficiency learners by compensating for linguistic gaps, while reflective scaffolding benefited higher-proficiency learners by activating metacognitive capabilities without imposing redundant support. These effects were most pronounced for originality, with language proficiency moderating the relationship between scaffolding type and creative outcomes. Qualitative findings from interviews and artifact analysis corroborated these patterns. The findings extend CLT and ATI theory into GenAI-enhanced learning environments, demonstrating that optimal scaffolding depends on alignment between learner capabilities and treatment characteristics. Practical implications emphasize teachers’ evolving role as "scaffolding diagnosticians" who orchestrate adaptive AI support as learners progress, contributing a differentiated framework for implementing GenAI in PjBL.