AI in Measurement-Based Learning: A Challenge for Assessment, an Opportunity for Tutoring
Sami Suhonen · 2025
The rapid advancement of artificial intelligence (AI) has introduced significant challenges to traditional assessment methods in engineering education. While AI-powered tools, such as ChatGPT, can enhance learning by assisting with data analysis and problem-solving, they also raise concerns regarding academic integrity and students' conceptual understanding. This study explores how AI interacts with measurement-based physics assignments, focusing on whether AI can solve these tasks independently and how it can be effectively integrated as a learning tool rather than a substitute for student effort. Five measurement assignments, covering topics such as acceleration analysis, friction coefficients, specific heat capacity, discharge coefficients, and muzzle velocity estimation, were given to ChatGPT under the same conditions as human students. The AI's performance was evaluated based on accuracy, problem-solving methodology, and ability to visualize results. Findings indicate that ChatGPT can successfully apply theoretical models and provide structured solutions and even exemplary data and graphs to many measurement assignments. Sometimes it fails to identify key experimental limitations, such as real-world heat loss and sensor calibration issues. Nevertheless, it clearly outperforms many bachelor's level engineering students. The results highlight the need for AI-resilient assessment and grading methods, where students engage in hands-on data collection, critical thinking, and peer discussions to ensure deeper learning.