Learning Together: Modeling the Process of Student-AI Interactions When Generating Learning Resources

Abstract

Introducing Large Language Models (LLMs) into the classroom in an effective way implies that students must acquire the necessary AI literacy skills, such as prompt engineering. This proves challenging, given that the knowledge base around effective prompting is still developing. This paper reports on a case study in which students make use of an LLM to generate personalized learning materials in a social network analysis course. We analyze students’ qualitatively coded interactions with the LLM and their evolution throughout the course using Epistemic Network Analysis (ENA). Our findings show that students evolved their prompting strategies throughout the course, providing more contextual information in their initial prompts by the final assignment. Specifically, high achievers were more likely to add contextual information in initial prompts and focused on refining the output rather than engaging in conversational exchanges with the LLM. They were also more likely to use polite language in their interactions. Our results highlight the need for further research and training in effective prompting to take full advantage of the potential of LLMs in education and to improve student-AI collaboration in academic settings. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.