Inquiry-based learning in STEM education: the impact of generative AI-based Chatbots on primary school students’ problem posing ability in science

Problem posing serves as the starting point for inquiry-based learning in science, technology, engineering, and mathematics (STEM) education. However, it has not received sufficient attention in primary science education practices. Although the widespread adoption of generative artificial intelligence (GenAI) presents new opportunities to improve this situation, empirical evidence regarding how it facilitates problem posing in inquiry-based learning remains limited. This study employed a quasi-experimental design to systematically compare the effects of GenAI-based chatbots and search engines on primary school students’ problem posing ability in science. A total of 97 third-grade students from a primary school in central China were randomly assigned by class to the experimental group ( N = 48, using a GenAI-based chatbot) and the control group ( N = 49, using a search engine). Following three rounds of progressive problem-based inquiry learning, students’ problem posing ability, cognitive network structures, and learning experiences were analyzed and evaluated. Results showed that the use of chatbots significantly improved problem quality ( t = 2.47, p = 0.015) and overall problem posing ability ( t = 3.07, p = 0.003). Epistemic network analysis further showed that the experimental group developed a more integrated network structure, whereas the control group exhibited a pattern of local clustering. In addition, the experimental group reported lower cognitive load and higher technology acceptance. Overall, GenAI-based chatbots significantly outperformed search engines in facilitating problem posing. This study reveals the potential of GenAI in STEM education, providing empirical evidence for introducing GenAI tools into inquiry-based learning in primary science classrooms.