211 computer science students use LLMs in very different ways depending on the subject, and almost always check

Algorithms and web development generate the most queries, whilst software engineering generates the fewest; the majority of students view the model as an aid and combine several verification methods.

Researchers analyzed the reflections written by 211 undergraduate students after each assignment in a problem-solving course designed to support the responsible use of large language models. The data covers seven modules corresponding to different areas of computer science, with indicators of the number of queries, how students perceive the role of the model, and their verification practices.

The type of duty shapes usage

Adoption varies greatly from year to year: it is highest in algorithmics and web development, and lowest in software engineering. Overall, students view the model as a back-up tool rather than an authoritative source, and answer verification is common across all domains, with most combining multiple strategies.

The way to check also depends on the context: tests are preferred for structured tasks, Web research for open problems. For the authors, even with the same pedagogy, it is the characteristics of each assignment that determine the students' relationship to AI, hence the interest in adapting teaching by sub-discipline.