Cognitive Boundaries in AI-Assisted Learning for Adolescents: Support, Substitution, and Guidance

Generative artificial intelligence (AI) is increasingly integrated into adolescent learning scenarios, enhancing the efficiency of task completion while raising concerns about the potential weakening of learners’ cognitive boundaries. Drawing on distributed cognition, cognitive load, and self-regulated learning theories, this study clarifies the distinction between cognitive support and cognitive substitution based on whether learners retain cognitive agency. Cognitive boundaries are conceptualized along three dimensions: cognitive stages, learner responsibility, and cognitive depth. Due to their critical period of cognitive development, adolescents’ cognitive boundaries are particularly vulnerable, which manifest through four mechanisms: premature termination of problem construction, outsourcing of cognitive processes, implicit weakening of metacognitive monitoring, and structural shifts in learning agency. To address these challenges, the study proposes guidance strategies across three levels: instructional design, teacher-student interaction, and institutional policy. These strategies aim to balance the empowerment offered by AI with the preservation of learners’ cognitive development, supporting autonomous, reflective, and effective human-AI collaborative learning. By elucidating both the opportunities and risks associated with in-depth AI integration, this study provides theoretical and practical insights for designing adolescent learning experiences that maintain cognitive engagement and promote meaningful learning outcomes.