Instructor Evaluation of Meta Smart Glasses: Assessing AI Accuracy, Explanation Quality, and Misconception Detection in Physics

This study investigates how effectively the Llama2-based AI embedded in Ray-Ban Meta Smart Glasses solves upper-level high school physics problems, classifies problem types, and addresses common misconceptions. Fifteen experienced physics instructors tested the AI’s performance behind the Meta smart glasses on three representative problems — projectile motion, Newtonian equilibrium, and energy conservation. We evaluated the AI on (1) accuracy, (2) explanation quality, (3) misconception handling, and (4) usability in a hands-free format. While the system demonstrated proficiency in recognizing problem types and generating fundamental quizzes, it showed inconsistencies in multistep calculations and occasional misinterpretation of prompts. This study does not assess students’ conceptual learning outcomes; rather, it focuses on instructors’ assessments of AI-generated solutions. Findings indicate that wearable AI effectively classifies physics problems and supports basic conceptual tasks, yet exhibits limitations in multistep problem-solving and formula application. The variability in instructor evaluations highlights the need for improved speech recognition, refined error-detection algorithms, and clearer multistep reasoning processes. Future research should include larger-scale studies directly measuring student conceptual change.