Using Artificial Intelligence for Assessment Support in Introductory Physics

Formative and summative assessments are critical components of effective physics teaching and learning. Historically, providing meaningful feedback on assessment types beyond closed-response formats (e.g., multiple choice or ranking questions), short-answer tasks (e.g., numerical responses or symbolic expressions), or specialized formats (e.g., pre-programmed simulations or Jupyter Notebooks) required human intervention. This talk presents experiences from ETH Zurich, where artificial intelligence, in combination with the open-source tool Ethel, has been used to offer feedback on handwritten homework solutions involving formulas and derivations in large introductory physics courses. AI has also been employed to assist in grading extensive handwritten exams. Emphasis is placed on providing confidence measures for AI-based judgments, fostering effective human-AI collaboration in the grading process. Additionally, early experiences with the automated generation of interactive assessment problems from course materials will be shared.