This study investigates the efficacy of Self-Explanation Prompts (SEPs) in enhancing problem-solving performance and metacognitive accuracy within STEM education, while simultaneously offering a comparative analysis of human versus artificial cognition. Grounded in the theoretical frameworks of Metacognition and Self-Regulated Learning (SRL), the research employs a quasi-experimental design with a diverse sample (N= 150, ages 10–50) divided into a SEP intervention group and a control group. Results indicate that structured reflective prompting significantly improves problem-solving accuracy and metacognitive calibration (Gamma correlation). Furthermore, the study contrasts human cognitive responses with those of three leading Large Language Models (ChatGPT, Perplexity, and Gemini). Findings reveal a fundamental divergence: while AI models excel at logical pattern matching, they lack the embodied, emotional, and contextual reasoning such as the intuitive understanding of physics or emotional pragmatics that characterises human thought. The study concludes that SEPs are essential for cultivating the self-aware, adaptive expertise that distinguishes human intelligence from algorithmic data processing.
Panya Samtani (Fri,) studied this question.