Pedagogical Scaffolding for Generative AI-Supported Self-Directed Learning: A Qualitative Study of the CRAFT Framework

Abstract

This study investigates how generative Artificial Intelligence (AI) can scaffold self-directed learning (SDL) in secondary classrooms. Focusing on the interaction between teacher facilitation and AI affordances, it examines and refines the Content–Response–AI&peers–Fusion–reflecT (CRAFT) framework as a practitioner-informed pedagogical scaffold for operationalizing aspects of the Process dimension of the Person–Process–Context (PPC) model. The study addresses a persistent gap between SDL theory and classroom practice by examining how learner agency can be cultivated within AI-supported learning environments. Adopting a qualitative multi-case study design, the research analyzed instructional practices, lesson artifacts, and reflective accounts from six experienced educators from a secondary school in Singapore across one academic year. Data were analyzed using a theoretically informed, hybrid deductive–inductive thematic analysis, enabling cross-case examination and iterative refinement of CRAFT in relation to the PPC model. Findings indicate that while generative AI supports goal setting, idea generation, and iterative feedback, the metacognitive and reflective dimensions of SDL remain underdeveloped without deliberate teacher mediation. Teachers remain critical orchestrators of AI and peer interactions, supporting disciplinary sense-making and learner autonomy rather than positioning AI as a standalone instructional aid. Overall, this study integrates a practitioner-informed and empirically refined pedagogical scaffold with an established theoretical model, offering a theoretically grounded, context-sensitive pedagogical guide for designing learning experiences that support opportunities for student agency.