Optimizing the Human-AI Learning Eco-system: Evaluating an AI-Scaffolded Andragogical Framework and Student Dynamics in Effective Classroom Practices

Abstract

In the current digital era, traditional teaching methods struggle to integrate evolving AI technologies with learning frameworks to meet the demands of the labour market. To satisfy the needs of current youth and future leaders, a mixed method quasi-experiment was conducted in a basic accountancy module at undergraduate level. The goal is to identify effective classroom practices under AI-integrated andragogical learning framework and improve student performance in areas of cognitive skills, social skills and efficient productivity. It evaluated an AI-scaffolded, andragogical learning framework with progressive difficulty levels across six student groups (A–F). Results: The approach yielded a 98.06% pass rate among 103 undergraduates, proving that class size had no impact and prior background did not prevent mastery. Variance analysis confirmed that assignment scores accurately reflected exam performance. Stronger academic groups adapted (groups D-F) quickest. Group D internalized the framework rapidly, while Group E used AI to overcome social-cognition struggles. Mediocre academic groups (A, B and C) faced wider performance variations as difficulty escalated, requiring more of the teacher’s support. Notably, weaker students relied entirely on AI, whereas stronger students cross-referenced AI data. Student engagement was exceptional; with over 72% of students completing all aggressive demands of ten team assignments of the module. Peer collaborations surged past 99% for groups A, B and C, when difficulty level escalated. However, Groups D and F reflected slower progress in developing social skills; and in only Group E students reverted to isolation under progressive difficulty. Ultimately, student satisfaction exceeded 90%, attendance rates remained relatively high.