Artificial intelligence in science education: global insights and future directions

relationships in learning, seeing AI as a co-participant that co-designs assessments for measuring students' usable knowledge in science education (Li et al., 2025).The global shift in science education toward helping learners use knowledge rather than merely recall it (Pellegrino & Hilton, 2012) makes the exploration of AI's role particularly timely.In the past decade, scholars have leveraged machine learning to automatically score student artifacts (e.g., Liu et al., 2016;Zhai et al., 2022), generate formative feedback for teachers (e.g., He et al., 2024b), and support students' model-based reasoning and explanation construction (e.g., He et al., 2024a;Zhu et al., 2020).These developments align closely with the core vision of the Next Generation Science Standards (NGSS, NGSS Lead States, 2013) and similar frameworks worldwide (e.g., Chinese Ministry of Education, 2017; Organisation for Economic Co-operation and Development, 2023), which emphasize the integration of disciplinary and interdisciplinary core ideas and scientific and engineering practices.However, the rapid acceleration of generative AI, especially systems such as ChatGPT, Gemini, and Claude, has dramatically expanded both the possibilities and uncertainties surrounding AI's use in STEM Education.Across contexts, teachers and researchers are grappling with several fundamental questions: How can AI be used to promote usable knowledge?In what ways can teachers leverage AI to interpret student thinking and provide timely, actionable, and constructive feedback?How do we ensure that AI-enhanced learning environments remain equitable, transparent, and aligned with human values?And perhaps most importantly, how do we prepare the next generation of teachers and students to engage with AI critically and responsibly?These questions demand not only technical innovation but also