AI-Powered Meta-Reflection for Self-Learning in STEM: A Middle School Case Study on Smart Home Energy Management

Emerging generative AI tools offer new opportunities to enhance students' metacognitive skills and self-regulated learning through guided reflection. This paper presents a case study on integrating AI-powered meta-reflection into a middle school STEM curriculum. We focus on a project-based learning (PBL) unit where students design a simple Smart Home Energy Management System, learning basic AI concepts, data analysis, and ethical implications of AI-driven decisions. A mixed-methods experimental design compared an AI-enhanced reflection approach (using OpenAI's ChatGPT) with traditional reflection activities. Quantitative data (pre/post tests, surveys, AI interaction logs) and qualitative data (student interviews, reflection journals, teacher feedback) were analyzed. Results indicate that students who engaged in dynamic AI-driven reflection showed greater gains in problem-solving and conceptual understanding, and demonstrated deeper self-regulated learning behaviors compared to a control group using conventional reflection. The AI prompted learners to analyze failures and iterate on solutions, improving their ability to learn from mistakes. Qualitative analyses suggest the AI reflection partner encouraged more introspective, adaptive thinking and higher engagement. We discuss how real-time conversational prompts from AI can scaffold meta-reflection, making student thinking visible and actionable for both learners and teachers. Key challenges (e.g., avoiding over-reliance or “metacognitive laziness”) are addressed through careful design of AI prompts and teacher oversight. Recommendations are provided for scaling AI-driven meta-reflection in K-12 STEM, including aligning AI tools with curriculum goals and training educators to integrate AI reflection facilitation. The study contributes initial evidence that AI-powered meta-reflection can enhance self-learning skills in middle school STEM, preparing students for more independent and reflective learning in an AI-rich future.