A COGNITIVE-PRESERVATION FRAMEWORK FOR YOUNG LEARNERS IN THE AGE OF GENERATIVE AI

Abstract

Generative artificial intelligence (AI) can expand access to explanation, feedback, and personalized support, but it can also perform cognitive operations through which learning is built. This conceptual review examines how young learners can use generative AI without outsourcing retrieval, reasoning, argumentation, verification, and transfer. A large field experiment in high-school mathematics found that unrestricted GPT-4 assistance improved practice performance but reduced later independent performance, whereas pedagogical guardrails largely mitigated this effect. Other recent evidence is mainly correlational or preliminary and does not justify broad causal claims. Drawing on cognitive offloading, cognitive load theory, retrieval practice, generative learning, metacognition, and emerging AI studies, the article proposes a three-phase Cognitive-Preservation Framework: think before AI, think with AI, and think again without AI. The framework positions AI as scaffold, tutor, critic, and intellectual adversary while preserving productive mental effort.