Investigating the Effect of High School Students’ Interactions with ChatGPT on Metacognitive Activities in Mathematical Problem Solving

This study investigates how ChatGPT can support metacognitive processes during problem-solving activities. A class of 12 female students participated in two work sessions: in the first, the students interacted with ChatGPT without pre-training; in the second, the chatbot was trained with a structured prompt to act as an educational tutor. A qualitative analysis conducted according to Meijer and colleagues’ taxonomy of metacognitive activities showed that, with untrained chatbots, students mainly activate orientation and planning phases, while evaluation and elaboration rarely emerge. On the contrary, the structured prompt encourages engagement in all metacognitive activities, with longer reflection times and less linear sequencing. The study suggests that ChatGPT can support metacognitive activities when properly trained, highlighting the importance of prompt design and the role of the teacher as an orchestrator of interaction. Educational implications, limitations of the study, and directions for future research are discussed.