Using artificial intelligence in high school STEM education improves students' cognitive and emotional development.

Objective:
The study aimed to systematically review existing empirical research on how artificial intelligence is used in high school STEM education. The goal was to summarize the research focus, methods, outcomes, and limitations to understand how AI affects learning in this specific educational context.

Methods:
This was a systematic review that followed the PRISMA protocol. The researchers searched the Scopus database for peer-reviewed, empirical studies involving high school students and AI applications in STEM education. Only studies that met criteria—such as being in English, conducted at the high school level, and focused on STEM subjects—were included. They started with 188 documents and filtered down to 8 articles. A coding form was used to capture key characteristics of each study, including year, authors, location, participants, AI type, method, and findings. Two researchers reviewed and analyzed the studies, and discrepancies were resolved with a third reviewer. The agreement rate for coding was 92%.

Key Findings:

• Most studies used a single-group pre/post-test design; only two used control groups.
• Sample sizes ranged from 16 to over 10,000 students, with the largest study using machine learning to analyze mindset interventions.
• AI tools ranged from intelligent tutoring systems and dialogue systems to machine learning algorithms and neural networks.
• The studies focused on a wide range of outcomes including:
  • Increased understanding of science and physics concepts.
  • Improved computational thinking (CT) and problem-solving skills.
  • Higher confidence in discussing and applying AI.
  • Stronger interest in STEM careers.
  • Enhanced engagement and learning outcomes from adaptive tutoring systems.
• Almost all studies reported statistically significant positive effects on learning, whether cognitive (like grades and concept understanding) or affective (like motivation and interest).
• Pedagogical agents, interactive tools, and hands-on machine learning activities were especially effective in improving student self-efficacy and engagement.

Implications:
This review shows AI can be a meaningful part of STEM education in high schools. Its ability to personalize instruction, support hands-on learning, and improve student engagement makes it a promising tool for educators. The positive outcomes support broader adoption of AI technologies, not just to teach AI concepts, but to enhance how core STEM subjects are taught and learned.

Limitations:

• Small number of studies (only eight met all criteria).
• Most studies used small sample sizes, except one large-scale intervention.
• Heavy reliance on single-group designs, with limited use of control groups.
• Few studies followed students long-term, so the lasting effects of AI use in classrooms remain unclear.
• Lack of consistency in measured outcomes—each study targeted different skills or attributes, making comparisons difficult.
• Minimal discussion of ethical concerns or privacy in AI-based education.

Future Directions:

• More research with control or comparison groups is needed to clarify cause-and-effect relationships.
• Longitudinal studies could help understand the lasting impact of AI in classrooms.
• Future work should focus on cross-age curricula, standardized measures, and deeper integration of learning theory with AI applications.
• There’s a need to explore ethical implications and privacy concerns more thoroughly.
• Researchers should pay more attention to demographic-specific needs and outcomes, especially when tailoring AI tools for different student populations.

Title and Authors:
"A Systematic Review of Artificial Intelligence in High School STEM Education Research"
By Aigul I. Akhmetova, Damira M. Sovetkanova, Lyazzat K. Komekbayeva, Assan E. Abdrakhmanov, Daniyar Yessenuly, and Oral S. Serikova

Published On: March 21, 2025

Published By: EURASIA Journal of Mathematics, Science and Technology Education, Volume 21, Issue 4, Article em2623 DOI: https://doi.org/10.29333/ejmste/16222