Students’ Difficulties in Learning about AI from Teachers' Perspectives: Insights from an Action Research Approach in Compulsory CS Education

Abstract

The omnipresence of AI in everyday life highlights the importance of integrating learning about AI into K-12 computer science (CS) curricula to prepare students for the responsible use of this technology. However, empirical research on AI-specific teaching and learning processes in compulsory CS education is still emerging, with limited findings on students’ learning difficulties. This paper presents findings from a participatory action research study on students’ learning difficulties in compulsory CS lessons about AI in Year 11. The first iteration intentionally foregrounded teachers’ perspectives through data from lesson reflection protocols and semi-structured interviews. To complement these perspectives, a post-instruction learning assessment was carried out. Using a grounded theory-informed analysis, we derived 4 core categories of AI-specific difficulties: (1) evaluating whether real-world systems use AI approaches, (2) distinguishing between knowledge-based and data-based approaches, (3) evaluating AI models, and (4) applying AI concepts to real-world examples. Our analysis further revealed possible reasons for these difficulties, including students’ preconceptions, the lack of adequate tools, and the complexity of real-world AI systems. In a final workshop, the findings were collaboratively discussed, and implications for addressing the difficulties in future lessons were formulated. The study’s findings support the advancement of AI education and offer a foundation for future research on AI-specific teaching and learning processes.