Abstract:

The flipped classroom is an active-learning instructional model in which students independently prepare content-based materials prior to class, followed by in-depth, instructor-moderated peer discussion focused on critical analysis, argumentation, and scholarly dialogue. This approach is particularly effective in higher education settings requiring autonomy and meta-cognitive skills.

Since the public emergence of powerful large language models (LLMs) in late 2022, university students have increasingly utilised these tools for assignment preparation, generating polarised faculty views on academic integrity and pedagogical value. This exploratory study examines student perceptions and experiences when LLMs are transparently integrated within a flipped-classroom framework, focusing on preparation strategies, perceived learning benefits, and attitudes toward AI use in higher education.

An exploratory study was conducted in a master’s-level course on mathematics topics for students in mechatronics and mechanical engineering, involving 68 participants. Students were permitted to use either traditional academic materials, publicly available LLMs (e.g., ChatGPT, Claude, or Gemini), or a combination of both during preparation. Those employing AI tools were encouraged to document prompts, iterations, and final human-edited outputs.

In-class sessions comprised student presentations followed by instructor-led open discussions, during which participants reflected on their preparation experiences and compared the perceived effectiveness of book-only, AI-only, and combined approaches. These discussions evaluated content accuracy, depth of argumentation, source evaluation, originality, and rhetorical strength, while also exploring subjective perceptions of engagement and confidence across preparation methods.

Student experiences and perceived outcomes were further assessed via a post-course questionnaire (20 multiple-choice items and 5 open-ended questions) administered on the final day, addressing preparation strategies, understanding, discussion quality, and attitudes toward AI in education. Qualitative analysis of open-ended responses revealed two prominent themes: (1) novelty-driven engagement and motivational activation, as students reported increased curiosity, effort, and exploratory behaviour facilitated by AI; and (2) targeted inquiry and meta-cognitive scaffolding, with students highlighting iterative prompting to resolve specific knowledge gaps.

Findings indicate that students who used AI tools reported higher engagement and greater confidence in their understanding of the material. Informal class discussions suggested no perceived differences in presentation quality across preparation methods, though this was not formally assessed. When LLM use is purposeful, transparent, and embedded within a flipped-classroom pedagogy, students perceived that academic integrity was maintained while motivation was enhanced. These results support a reconceptualization of academic integrity in the AI era—from prohibition to pedagogically scaffolded human–AI collaboration in higher education.

Keywords:

Flipped classroom, large language models, academic integrity, human–AI collaboration, motivational engagement, active learning.