Abstract:

This study investigates how university students understand and use generative Artificial Intelligence (AI) in their learning processes, and explores the educational implications emerging from these practices. As generative AI rapidly becomes embedded in academic environments, universities face growing challenges related to academic integrity, cognitive reliance, and the need for new forms of AI literacy. To address these issues, this research focuses on two groups with contrasting learning contexts: first-year students enrolled in an introductory AI and data science course, and fourth-year students engaged in undergraduate thesis writing. These groups provide an opportunity to compare how generative AI is incorporated into different academic stages, from foundational learning to advanced research activities.

A multi-dimensional questionnaire was administered to examine students’ usage frequency, purposes of use, perceived benefits and drawbacks, anxieties, dissatisfaction, understanding of AI mechanisms, impact on learning behaviors, ethical concerns, and future intentions of use. Findings indicate that generative AI has already become a routine part of students’ academic work. However, substantial differences appear in how AI is used. First-year students primarily employ AI for clarifying concepts, checking their understanding of lectures, and obtaining explanations of unfamiliar terms. For them, generative AI functions as a supportive tool that assists basic comprehension and reduces learning barriers.

In contrast, fourth-year students use AI for tasks directly connected to research productivity, such as exploring thesis directions, organizing literature, drafting and revising academic writing, and refining arguments. For these advanced learners, AI is perceived as an intellectual partner that accelerates complex processes and enhances the quality of academic output. This contrast highlights how students’ reliance on AI evolves with academic maturity and task complexity.

The perceived benefits reported by both groups include improved understanding of difficult topics, increased learning efficiency, support for writing tasks, assistance with structuring ideas, and access to alternative viewpoints. At the same time, significant concerns were noted.

A notable finding is that fourth-year students demonstrate a higher understanding of how generative AI works, including awareness of its limitations. This suggests that technical literacy contributes to more responsible and critical use. Across both groups, students commonly requested explicit guidelines defining appropriate and inappropriate uses of AI for assignments and research.

Overall, the study suggests that generative AI can contribute meaningfully to learning when embedded within a well-designed instructional framework that emphasizes critical thinking, ethical awareness, and developmental appropriateness. The results underscore the importance of differentiated AI literacy education, clear institutional policies, and pedagogical approaches that position AI not as a replacement for human reasoning but as a cognitive partner. These insights provide a foundation for rethinking how generative AI should be integrated into higher education in ways that support student growth while preserving academic integrity.

Keywords:

Generative AI, higher education, comprehensive survey, learning behavior.