Generative Artificial Intelligence‐Assisted <scp>STEM</scp> Education: A Systematic Review and Future Research Agenda
Yifan Zhu, Haozhe Jiang, Luoxin Zhong
Abstract
ABSTRACT As generative artificial intelligence (GenAI) gains popularity, there is still a significant gap in understanding its potential to enhance STEM education. To address this gap, we conducted a systematic review of 55 studies through the lens of the five‐dimensional technology‐based learning framework (devices, domains, methods, participants and research issues). Following PRISMA guidelines, we extracted and coded each paper's GenAI tool, subject area, methodology, participant profile, sample size, and the reported benefits and barriers. The majority of articles focused on GenAI‐assisted science education, followed by technology education, while research on engineering, mathematics and integrated STEM learning remained relatively limited. Text‐based GenAI was the most commonly used tool, while image‐based, video‐based, music‐based and audio‐based GenAI were rarely featured. Most studies concentrated on higher STEM education, with limited research focusing on K‐12 STEM education. Quantitative methods were predominantly used to assess the effectiveness of GenAI. GenAI offered significant benefits in STEM education, positively influencing individual learners, teachers, and the learning and teaching process. However, GenAI also posed challenges for learners, including unreliability and inaccuracy, ethical and academic integrity concerns, dependence, reduced critical thinking skills and learning efficiency, and accessibility barriers. Teachers faced challenges like limited instructional support and the need to adopt more student‐centred approaches. Based on the results, we suggested several research priorities to further explore GenAI‐assisted STEM education. Additionally, we outlined key implications for learners and educators to fully leverage GenAI's benefits and overcome the obstacles.