Laura Kobel-Keller | Searcharxiv

arXiv2026

Typed Mathematical Text for On-screen Examinations

This paper discusses digital online mathematics examinations -- a discussion ranging from high school to university level examinations. In particular, we consider the nature of mathematical writing, what is distinctive about mathematical writing, and how mathematics can be typed into a machine. This includes a review of features of notation and layout unique to mathematics and a survey of current technology for typed mathematics, including LaTeX and contemporary proof-checkers such as Lean. Artificial intelligence has already been highly successful for optical character recognition, generating text from hand writing and is even increasingly applied to assess students' work itself. A human-editable text-based format in the middle is important, but neglected. The design of digital online mathematics examinations, which take this text as the source of truth for students' work, will have a profound effect on how mathematics is perceived and how mathematical activity is mediated. Moving examinations on-screen effectively is an important design challenge and responsibility to future generations. The challenge is to design software tools which support mathematics, that is tools which recede into the background and support the generation of mathematical work. We argue for a human-editable text-based format, which includes semantic elements, at the heart of the process.

math.HO↗

arXiv2024

AI-assisted Automated Short Answer Grading of Handwritten University Level Mathematics Exams

Effective and timely feedback in educational assessments is essential but labor-intensive, especially for complex tasks. Recent developments in automated feedback systems, ranging from deterministic response grading to the evaluation of semi-open and open-ended essays, have been facilitated by advances in machine learning. The emergence of pre-trained Large Language Models, such as GPT-4, offers promising new opportunities for efficiently processing diverse response types with minimal customization. This study evaluates the effectiveness of a pre-trained GPT-4 model in grading semi-open handwritten responses in a university-level mathematics exam. Our findings indicate that GPT-4 provides surprisingly reliable and cost-effective initial grading, subject to subsequent human verification. Future research should focus on refining grading rules and enhancing the extraction of handwritten responses to further leverage these technologies.

math.HO↗