Study Analyzes Student Question Types in AI Interactions Across CS2 Programming Tasks

A study examining 830 interactions between students in CS2 and generative AI systems revealed that a limited range of question types makes up most student inquiries. Posted on arXiv, the research utilized the Graesser et al. taxonomy to categorize these inquiries into 18 distinct types and employed a few-shot learning method for automatic classification. The data was sourced from two programming assignments. The researchers sought to explore the nature of questions students pose to AI and how these questions change during problem-solving and across different tasks. This research highlights the contradiction of students seldom asking questions in class while frequently using AI for their learning. The automated classification facilitated extensive analysis, uncovering a concentration of specific inquiry types.