Exams
Making reasoning-focused, handwritten exams practical at scale, while keeping educational judgment in human hands.
In an age when AI can readily produce answers, understanding how to reach them matters more than ever. Handwritten, open-ended exams give students room to develop derivations, construct arguments, and sketch their thinking - and to earn partial credit for sound reasoning even when a final answer is wrong. Our motivation is to preserve these possibilities in large classes, rather than letting the workload of scoring determine what we assess.
Our work has progressed from research studies in mathematics and the sciences to AI-assisted scoring in regular ETH examinations. We combine instructor-defined rubrics with AI support and human oversight, examining both the potential and the limitations of the approach. Instructors and students can veto individual scores, triggering independent human rescoring, while examiners retain responsibility for final grades. The aim is not simply faster scoring, but sustainable support for research-based assessment practices.
Additional information can be found in the author guide.
First Experiences
During ETH Zurich’s summer 2026 examination session, our AI-assisted workflow supported scoring of approximately 2,000 handwritten exams across six courses, covering around 40,000 pages and 28,559 problem-part scores. Of these scores, 90.8% were not vetoed and required no manual rescoring. The remaining 9.2% were vetoed by instructors or students and independently rescored by humans, without access to the AI’s original score or reasoning. Among vetoed scores, 54.8% remained unchanged, 34.8% increased and 10.3% decreased. Overall, nearly 96% of final scores matched the initial AI scores.
Publications
- external page Gerd Kortemeyer, Toward AI grading of student problem solutions in introductory physics: A feasibility study, Phys. Rev. Phys. Educ. Res. 19, 20163‐1 — 20163‐20 (2023)
- external page Gerd Kortemeyer, Performance of the Pre-Trained Large Language Model GPT-4 on Automated Short Answer Grading, Discover Artificial Intelligence 4, 47 (2024)
- external page Gerd Kortemeyer, Julian Nöhl, and Daria Onishchuk, Grading assistance for a handwritten thermodynamics exam using artificial intelligence: An exploratory study, Phys. Rev. Phys. Educ. Res. 20, 20144‐1 — 20144‐24 (2024)
- external page Gerd Kortemeyer and Julian Nöhl, Assessing confidence in AI-assisted grading of physics exams through psychometrics: An exploratory study, Phys. Rev. Phys. Educ. Res. 21, 010136‐1 — 010136‐24 (2025)
- external page Tianyi Liu, Julia Chatain, Laura Kobel-Keller, Gerd Kortemeyer, Thomas Willwacher, and Mrinmaya Sachan, AI-assisted automated short answer grading of handwritten university-level mathematics exam, Teaching Mathematics and its Applications 45, 84 — 105 (2026)
- external page Jan Cvengros and Gerd Kortemeyer, Assisting the grading of a handwritten general chemistry exam with artificial intelligence, Discover Artificial Intelligence 6, 931 (2026)
Contact:
Alina Yaroshchuk