The empowerment of educational reform and innovation through AI technology has become a topic of increasing interest in the field of education. The advent of AI technology has made comprehensive and in-depth teaching evaluation possible, serving as a significant driving force for efficient and precise teaching. There were few empirical studies on the application of high-quality precision teaching models in the field of compulsory education, and the learning difficulty of technology and the teaching burden on teachers have become significant factors hindering the use of technology to support education. This study analyzed teaching models from the perspectives of teachers’ teaching burdens and students’ learning obstacles, and was committed to relying on intelligent technology to construct a new precision teaching model, an educational diagnosis–feedback–intervention path that covered the entire teaching process, from the dimensions of teacher behavior, student behavior, and parent behavior, aiming to assist teachers in efficient teaching and students in personalized learning. This study was conducted with nine science classes, including about 540 people in the second year of high school at a Middle School in China; six classes were the intervention groups while the last three classes were control groups, and a survey of 19 teachers from the intervention classes was carried out. The results showed that this model can significantly improve students’ academic performance in science subjects, especially in mathematics and chemistry. It has increased the proportion of high-achieving students, reduced the proportion of low-achieving students, stimulated students’ self-directed learning ability, cultivated a positive attitude towards science learning, and explained the key points of using a precision teaching model in different disciplines. It has achieved a deep integration of education and technology, helping to increase the efficiency and reduce the burden of teaching.

Computational thinking, the thinking processes and skills for analyzing and developing computational solutions to solve real-world problems, has been increasingly cultivated through STEM (Science, Technology, Engineering, and Mathematics) programs that integrate programming and non-programming disciplines in authentic problem-solving tasks. To guide students in learning with complex programming tasks, existing studies have employed explicit instructional models (i.e., identifying problems, decomposing programming tasks, coding, and optimizing programs). However, these models are inadequate to support young learners, who often struggle to connect non-programming knowledge to programming processes and accomplish each phase of the complex task. This study proposed the ICDCO instructional model that features (a) incorporating interdisciplinary knowledge acquisition into the task process and (b) providing rules-of-thumb to guide each phase of the task. An action research study was conducted with 136 students from three primary schools in Mainland China and Macau. Results show that after taking a STEM course designed using the ICDCO instructional model, students significantly increased their overall self-perceived computational thinking skills and performed well in decomposing a complex problem. The impact of the approach was affected by class size and students’ educational level but not by gender. The findings suggest that the proposed approach can benefit students’ computational thinking skills and performance in decomposing a complex task. More support should be provided for students in large classes and at lower educational levels. Further research is needed to examine the effects of the ICDCO instructional model and the factors that influence its application in cultivating computational thinking in STEM courses.