Designing AI-powered Scaffolds to Foster Metacognition and Agency in EFL Writing

Abstract

Learners in English as a Foreign Language (EFL) writing often struggle with limited metacognitive strategy use, weak feedback literacy, and insufficient self-regulation. EFL writing requires the coordination of language knowledge, cognitive strategies, and metacognitive skills, which makes these challenges especially significant. While AI writing assistants offer personalized support, most emphasize performance optimization and error correction, which risks reducing learner agency and hindering metacognitive development. This research aims to create an AI-enhanced writing platform designed to foster both writing performance and learner agency. It integrates LLM-based metacognitive prompts, adjustable agency control, and analytics dashboard within a digital writing environment, guided by theories of self-regulated learning, structure and agency, and recursive writing process. Using a Design-Based Research approach, this research involves iterative design, classroom-based pilot studies, and mixed-method evaluation. It has potential contributions include empirical evidence for balanced AI support, theoretical expansion of agency framework into AI-enhanced writing, and practical design principles to enhance learner agency. Preliminary findings and future directions will be discussed in the workshop.