Designing Generative AI-Based Agents to Empower Collaborative Self-Learning in Higher Education

Abstract The emergence of Generative AI (GenAI) is rapidly transforming higher education by offering new opportunities for personalized learning and supporting students on their individual learning journeys. Despite GenAI’s potential to enhance student engagement and learning outcomes, many current GenAI-based learning tools remain weakly grounded in pedagogical frameworks necessary for meaningful educational impact. Addressing this gap, this paper investigates how to design GenAI-based agents to effectively support student motivation and learning processes by drawing on self-directed learning as a pedagogical kernel theory. By adopting a multi-cycle Design Science Research approach, prescriptive design knowledge is derived through qualitative interviews and literature in two design cycles. The resulting design principles are instantiated within a GenAI-based prototype for a university course and subsequently evaluated empirically by means of a user study and expert interviews. The study contributes a set of theoretically grounded and empirically informed design principles for GenAI-based learning agents that support self-directed learning while complementing rather than replacing students’ critical thinking. By articulating actionable design knowledge, this research supports the pedagogically informed integration of GenAI into higher education and contributes to the evolving discourse on GenAI-enabled learning in the context of Education 4.0.