Many universities have incorporated computational thinking and artificial intelligence (AI) courses into their general education curricula for first-year students. In parallel, efforts have been made to diversify instructional approaches to accommodate students’ varying levels of academic background. However, high-level courses often require substantial prerequisite knowledge and present complex content, resulting in significant cognitive burden. Particularly in stratified AI education models, students enrolled in advanced-level courses frequently experience heightened cognitive overload.
This study examines a machine learning course offered as part of the general education curriculum, aiming to assess students’ cognitive burden, the perceived coherence of course components, and their responses to instructional strategies, with the goal of deriving practical implications for course improvement. Specifically, student responses were structured around differences in perceived difficulty across academic disciplines, preferences for instructional methods (lecture, practice, and project), and experiences of conceptual application during project work.
Employing a case study approach, the research draws on survey responses from 135 students, open-ended feedback from course evaluations, and narrative data from project reports. Findings indicate that perceptions of course difficulty and learning burden differ based on academic background and prior experience, with cognitive overload being most pronounced in the artificial neural networks unit. Additionally, weak integration between practice and project phases and limited applicability of practical exercises to real-world problems were found to exacerbate students’ cognitive burden.
These results underscore the need for liberal arts-oriented machine learning courses to ensure sequential integration of theory, practice, and projects while addressing students’ cognitive load. The study concludes by recommending instructional strategies aimed at enhancing learning effectiveness and reducing cognitive overload, including reinforcement of iterative cycles of concept explanation and practice, incorporation of public data-based assignments, and increased instructional flexibility responsive to student feedback.