Impact of AI-Assisted Secondary Science Learning and Metacognitive Strategies on Student Performance

Abstract

This action research project compared 9th-grade students who rely on traditional study methods such as note-taking, retrieval practice, and concept mapping, with students who were taught how to use LLM-based tools to set study goals, self-monitoring and self-regulating habits. Artificial intelligence (AI) software can be valuable adjunct in science classrooms to promote personalized learning and to support metacognitive skill development. Large language models (LLM) can assist students in grasping concepts, generating questions, and optimizing study planning. Over a twelve-week treatment, the AI cohort was given three LLM models to choose from and then shown how to train these programs to help the student study the unit materials. The traditional cohort was given explicit training and structure on non-ai study methods. Most of the existing research in the literature has been conducted on university students, with few examples of high school science students. This project bridges this gap by implementing AI tools in the classroom and showing how LLM models can support self-regulation learning, time management and selftesting for early high school students. The results showed that AI-assisted helped a lower-starting AI cohort nearly match the final performance of a higher-starting traditional cohort. AI-based self-testing showed more rapid increases in confidence, perceived preparedness, and active recall routines. AI-assisted studying helped a lower starting cohort narrow the achievement gap in optimizing study techniques. These findings can inform effective strategies for implementing study skills for 9th grade students.