Let’s put an end to multiple choice!

This project develops an AI-assisted assessment system building on prior research. The system will be implemented and evaluated using real student data and feedback.

Key data

Description

Assessing a student’s knowledge and giving feedback is perhaps just as or even more crucial as conveying new information. It is a fundamental part of learning, and yet, we argue, it is one of the few areas that has deteriorated because of digitalisation rather than improved. Exams are increasingly designed with simple automated grading in mind such as multiple-choice or one-word answers. Open-ended questions, where students have to show their thinking, are not only more accurate measures of a student’s understanding but also boost learning and retention. However, they can be prohibitively time-intensive to grade.

This project will evaluate automated short-answer grading and feedback approaches on a blended learning ZHAW course for bachelor students in Fall’25 (~150 students). We will use an LLM-based method to grade student answers and give feedback. By allowing students to verify the grading approach themselves, we aim to enhance their satisfaction and build trust in the system.