From Gradual Release of Responsibility to Gradual Release of Technology: A Case for the Staged Use of AI in Formal Education

Abstract

Debate over the role of generative artificial intelligence in formal education is marked by sharply divergent claims about its effects on student learning. Much of this disagreement reflects a failure to distinguish immediate performance gains from durable learning. This article addresses a central gap in the literature by focusing on when and how students should be permitted to rely on AI. It introduces the Gradual Release of AI-Technology (GRAIT) model, extending Pearson and Gallagher’s gradual release of responsibility to the staged use of cognitively substantive AI-technologies in formal education. The model distinguishes gradual release of responsibility, which concerns who performs the cognitive work, from gradual release of technology, which concerns when and how that work may be scaffolded, augmented, or offloaded. Three phases are proposed: independent cognition, where foundational knowledge and skill are developed without AI substitution; shared (entangled) cognition, where AI is introduced through constrained, teacher-guided interaction; and offloaded cognition, where learners strategically delegate and reallocate aspects of thinking once risks to learning are minimal and opportunities emerge elsewhere. Progression depends on demonstrated knowledge, skill proficiency, and epistemic readiness. The central question is not whether students should use AI, but who should use it, when, how, and under what conditions it should be embraced or avoided.