Adaptive intelligent tutoring systems for STEM education: analysis of the learning impact and effectiveness of personalized feedback

Abstract Teaching Science, Technology, Engineering, and Mathematics (STEM) disciplines faces significant challenges due to the increasing complexity of content and the diversity in students’ knowledge levels. Traditional teaching methods, characterized by their standardized approach, fail to adapt to individual needs, limiting learning potential. This problem is particularly evident in mathematics, physics, and programming, where mastering critical concepts requires a logical and sequential progression. This study proposes an artificial intelligence-based intelligent tutoring system to provide a real-time personalized and adaptive learning experience. The system integrates advanced deep learning and natural language processing models, allowing for targeted feedback and dynamic adjustment of learning trajectories. The results show significant improvements in several key metrics. The experimental group achieved an average precision of 85% in programming and 78% in mathematics, significantly outperforming the control group. Furthermore, a linear regression model identified a positive correlation between the time of interaction with the system and the rate of progress in fundamental concepts. Student perceptions were also highly positive, with 80% appreciating the usefulness of adaptive feedback. These findings identify the system’s potential to transform STEM teaching, address the lack of personalization, and improve learning in various educational settings.