This study proposes an indicator system for evaluating AI-assisted learning in higher education, combining evidence-based indicator development with expert-validated weighting. First, we review recent studies to extract candidate indicators and organize them into coherent dimensions. Next, a Delphi session with domain experts refines the second-order indicators and produces a measurable, non-redundant, implementation-ready index system. To capture interdependencies among indicators, we apply a hybrid Decision-Making Trial and Evaluation Laboratory–Analytic Network Process (DEMATEL–ANP, DANP) approach to derive global indicator weights. The framework is empirically illustrated through a course-level application to examine its decision usefulness, interpretability, and face validity based on expert evaluations and structured feedback from academic staff. The results indicate that pedagogical content quality, adaptivity (especially difficulty adjustment), formative feedback quality, and learner engagement act as key drivers in the evaluation network, while ethics-related indicators operate primarily as enabling constraints. The proposed framework provides a transparent and scalable tool for quality assurance in AI-assisted higher education, supporting instructional design, accreditation reporting, and continuous improvement.
Ilieva et al. (Sun,) studied this question.