Higher Education Students' Perception of AI for Self-Learning: A Preliminary Study

This work proposes an experiment to compare the subjective experience of higher education students while learning with AI tools rather than the 'traditional' way. To this end, 16 Bachelor students in Computer Science with no previous knowledge of Entity-Relationship (ER) diagrams were asked to acquire the knowledge necessary to create an ER diagram, either assisted by a GenAl tool (ChatGPT or Gemini) or a set of a Database course slides (describing ER modEIS and how to draft them). The participants were randomly divided into groups of 2 or 3 people, then each group was assigned to a GenAl tool or slides. After 4 hours, the students were asked to provide the sketched ER diagram and answer subjective questionnaires assessing the quality of information provided by the knowledge acquisition tool they used. Findings indicate that students working with GenA Is modelled the best diagrams, although they judged the results provided by GenAl as not accessible and less reliable than slides. Students were the most critical in the case of Gemini. On the contrary, students with slides deemed the didactical materials trustworthy but time-consuming; this hindered their ability to deliver a complete ER diagram. In general, these preliminary outcomes showed that students judged GenAls as a good support to be included in their learning toolkit. Instead, when self-learning a new topic, they indicated the traditional course slides as the best methodology.