Large Language Models in Mathematics Education: Practitioners as Pedagogical Prompt Engineers

Abstract

Effective classroom integration of large language models (LLMs) requires domain-specific Prompt Engineering (PE) competence among teachers and students. Existing PE literature remains largely technical or cross-disciplinary, offering limited guidance for mathematics education, where LLMs continue to struggle with mathematical reasoning. This paper proposes a pedagogically grounded framework that positions educators as pedagogical prompt engineers who scaffold students’ interactions with LLMs through discipline-specific prompting practices. The framework provides structured classroom exemplars to enhance mathematical reasoning, conceptual understanding, and critical evaluation of AI-generated outputs. By fostering critical AI literacy and human-centered inquiry, it supports deeper, reflective, and ethical engagement with intelligent technologies.