Comparative Effectiveness of AI Tools in Enhancing Feedback Quality and Efficiency in Undergraduate Mathematics Assessments

Abstract Book of the 10th World Conference on Research in Teaching and Education

Year: 2026

[PDF]

Comparative Effectiveness of AI Tools in Enhancing Feedback Quality and Efficiency in Undergraduate Mathematics Assessments

Dr. Shun Ling Chiang

ABSTRACT:

Students in the Department of Mathematics at Hong Kong Baptist University often report that feedback on assignments is too brief and unhelpful. Teaching assistants, facing heavy time pressure when marking large classes, tend to rely on generic comments, leaving students unclear about their errors and misconceptions.
This Teaching Development Grant project examined whether AI could improve the situation. Two tools — Gradescope by Turnitin (strong in recognising handwritten mathematics and grouping similar answers) and a specialised AI Grading Assistant (designed for symbolic content and rubric-aligned marking) — were tested against traditional manual grading.
Over two semesters, the same assessments were graded using all three methods in randomised order. Grading time, inter-rater reliability, feedback relevance, error detection, and stakeholder perceptions were measured through surveys and focus groups. Preliminary findings from Semester 1 and full results from Semester 2 will be presented at the conference.
The study demonstrates a practical and scalable approach to integrating AI into mathematics assessment while preserving the human element of teaching. The findings and recommendations offer useful guidance for other departments and institutions seeking to enhance feedback practices.

Keywords: AI grading; Mathematics Education; Hand-written Mathematics; Higher Education; Teaching Efficiency