Generative AI tools such as ChatGPT have quickly become part of everyday learning. University students use them to solve mathematical problems, improve their writing, ask for explanations, and receive feedback. While many studies have examined students’ attitudes toward these tools or how often they use them, we still know relatively little about how .students actually interact with AI during authentic learning
In this study, we analyzed more than 15,000 interactions between 821 university students and generative AI across six university courses, including mathematics, engineering, English, and social sciences. We developed a framework that characterizes each interaction according to the student’s cognitive goal and the focus of the conversation. Our findings show that student–AI interactions are not random. Instead, they follow recurring patterns that differ across learning contexts. By understanding how students naturally use AI during coursework, this research provides a foundation for designing AI-supported learning environments and for studying how these interactions relate to learning in the future.
Link to the full article: https://arxiv.org/pdf/2606.29442
Authors: Taelin Karidi, Ofra Amir, Ido Roll
Research group’s website: https://roll.net.technion.ac.il/
