FROM GENERATIVE AI TO AGENTIC LEARNING SYSTEMS: A HUMANAI COLLABORATION FRAMEWORK FOR THE FUTURE OF DISTANCE HIGHER EDUCATION

Аннотация

Generative AI entered universities as a tool that students consult on demand. The next generation of systems, AI agents that plan, remember and act over several steps, will take over parts of teaching, and distance education will feel this first. This article asks how responsibility should be divided between people and such agents in distance higher education. Following design science methodology, it combines an integrative review of experimental and survey evidence published in 2019-2026, regulatory documents and enrolment data for Uzbekistan. The evidence points in one direction: the same models raise or lower learning depending on how they are constrained. An AI tutor built around expert-written scaffolding produced learning gains 0.73 to 1.3 standard deviations larger than active learning in class, while unrestricted GPT-4 access raised practice scores by 48% but lowered later exam scores by 17%. Student use of AI is close to universal (92-95% in UK surveys, 60-72% in Uzbek samples), yet only about a third of students report receiving institutional training. In Uzbekistan, correspondence and distance students account for 32% of 1.85 million students, with 36 students per teacher. The proposed framework defines five levels of AI agency, from tool to orchestrating agent, and assigns each distance learning task a maximum level using three criteria: the stakes of the decision, the reversibility of errors and the strength of evidence. Tutoring and formative feedback may reach the tutor agent level under teacher oversight, whereas grading, progression and admission remain human decisions. Recommendations for universities and national AI policy follow.