Open-ended questions in exams are a vital tool for assessing students' knowledge, critical thinking, and problem-solving abilities. Unlike multiple-choice questions, they require students to construct their own responses, providing a more nuanced evaluation of their understanding. However, grading these questions consistently poses a significant challenge.

How can we ensure fairness and consistency in grading while managing the substantial workload involved? For lecturers, the workload represents the most significant challenge in grading open-ended questions, particularly in large classes. Secondary challenges include ensuring fairness and consistency, as human graders can be influenced by biases and mental fatigue, leading to potential inconsistencies in scoring. These challenges often discourage the use of open-ended questions, despite their pedagogical benefits.

Enter AI-assisted grading — a solution designed to augment, not replace, human capabilities without replacing them. Our platform leverages insights from literature, the latest advancements in AI technology, and input from lecturers at UZH to support graders throughout the evaluation process, effectively addressing these challenges. It has been tested across diverse subjects, including mathematics, law, and political science, providing valuable insights and practical benefits.

The findings indicate that AI assistance offers valuable support in grading, enhancing workflow with fea-tures like text highlighting and reducing graders' workload. The workflow and process ensure fairness and consistency, as many sources of inconsistency can be addressed through the use of software, such as rubrics. Although AI's accuracy remains too low for full automation and is often more lenient than human graders, it still provides significant support.

This website offers a comprehensive overview of our research, the development of the grading tool and its evaluation, as well as insights into prompting strategies and further lessons for educators. Dive deeper to discover how AI can revolutionize exam grading, offering a fairer, more efficient, and insightful assessment process. Are you ready to explore the future of grading?

We sincerely thank our collaborators and sponsors, namely Swissuniversities for funding the development of this use case as part of the P-8/DISK4U project; the University of Zurich and the Department of Finance — Teaching Center for sponsoring the development of the tool.