From Generation to Autonomy: A Framework for Understanding Generative, Agent, and Agentic AI in Higher Education

Abstract

The rapid evolution of artificial intelligence (AI) in higher education has moved beyond content generation toward increasingly autonomous systems, necessitating a clearer conceptual understanding of their roles and implications. This paper introduces the Generation-to-Autonomy (GtA) Framework, which positions Generative AI, Agent AI, and Agentic AI along a multidimensional continuum defined by autonomy, control, and intelligence. Drawing on a functional definition of educational intelligence, the framework distinguishes AI systems based on their capacity for content generation, task execution, and self-directed decision-making. The study provides a structured analysis of how Generative AI enhances cognitive support yet raises challenges related to academic integrity and epistemic reliability; how Agent AI enables task-oriented automation and human–AI collaboration while introducing risks of accountability diffusion and task misspecification; and how Agentic AI represents a paradigm shift toward autonomous educational systems, raising complex ethical, governance, and control concerns. By synthesizing these categories into a unified framework, the paper contributes theoretically by clarifying the evolving human–AI relationship—from tool use to partnership and delegation—and practically by offering guidance for curriculum design, assessment transformation, and differentiated governance in higher education. The findings suggest that as AI systems progress toward greater autonomy, educational institutions must adopt nuanced, capability-sensitive strategies that balance innovation with ethical responsibility. The GtA Framework provides a foundation for informed decision-making and future research in AI-integrated education.