A survey of 313 students, supplemented by 18 interviews, describes very uneven learning paths, and a skill that resists: knowing how to evaluate what AI produces.
How do students perceive the evolution of their generative AI skills over time, and what do they get out of it for their studies? Lei Fan and Fangxue Liu analyzed 313 responses to a retrospective questionnaire and conducted 18 interviews with students from several disciplines, cross-referencing statistics, structural equation modeling, and thematic analysis.
Heterogeneous trajectories
Three profiles dominate: 34.2% of students describe rapid progress followed by a plateau, 28.4% a steady improvement and 21.1% a slow start and then an acceleration. Progress is also uneven depending on the skills: judging the quality of the AI's answers is perceived as the most difficult to acquire. It is the homework that most often serves as a learning engine.
The interviews show increasingly precise instructions, adapted verifications and more selective delegation, but also fixed routines, a form of dependence, skipped reasoning, or progress that is partly due to the improvement of the tools themselves. Academic benefits are mainly reported for homework, much less for exams, and are reflected, according to the statistical model, through the transformation of the way homework is worked.
Read the full article on the original source
Go to the source →