Teachers’ Perceptions of Using Generative AI Mathematics Problems in Online Curricula

• In Event: Designing AI-Powered Personalized Educational Tools by Incorporating Teachers’ Perspectives

Abstract

Objectives

Generative AI (GenAI) is being used in mathematics education to allow curricula to integrate tasks adapted to students’ interests (Einarsson et al., 2024; Norberg et al., 2024). Context personalization is the practice of adapting to students’ interests in popular culture areas like sports or music, or career areas like nursing (Walkington et al., 2024). Here we examine how teachers conceptualize the opportunities and challenges of personalization through GenAI.

Theoretical Framework

Personalization can trigger students’ interest (Hidi & Renninger, 2006; Renninger & Hidi, 2022), while also leveraging students’ real-world knowledge, integrating funds of identity (Esteban-Guitart, 2021). One consideration is the depth of connections to students’ interests (Walkington & Bernacki, 2014), and whether the numbers in problems should change when problems are personalized. Keeping the numbers the same is easiest but may compromise authenticity. While research has begun to explore AI-powered personalization, little research has examined how teachers are thinking about this approach. Our research questions are:
1)What do teachers see as the affordances and constraints of GenAI personalized problems?
2)What design recommendations do teachers cite for implementing GenAI personalization?

Methods

Interviews were conducted with n=11 Algebra teachers (5 male, 6 female) who taught in urban schools in the United States. The teachers were racially diverse and had an average of 14.5 years teaching experience. Interviews were conducted via Zoom and recorded.

Data Sources

The teachers were shown mathematics problems situated in interest areas such as social media and sports, written by Chat-GPT. They were asked about a hypothetical curriculum feature where students would be able to toggle between different problem contexts for a given math problem. Responses were coded using open coding techniques (Saldaña, 2021), with codes developed in a bottom-up manner.

Findings

Teachers saw GenAI personalized problems as a way to draw upon students’ real-world knowledge (8 teachers) and activate interest (8 teachers): “If it's talking about a place, thing, or situation that they're actually familiar with, that they've actually had hands-on experience with, of course, it's gonna be a little bit easier for them.” Constraints were that students would still lack key fundamental mathematics knowledge (4 teachers) or motivation (5 teachers). With respect to design, 4 teachers were okay with the LLM changing the numbers, 2 said the numbers could change if it was a different problem, 2 said to only change numbers when necessary, 2 said to not to change the numbers, and 1 said it did not matter. The primary concerns the teachers brought up with changing numbers were being able to assist students and difficulties with collaboration.

Significance

Mathematics teachers show some optimism about using GenAI to personalize problems to students’ interests. However, teachers accentuate that the issues they face are much larger than what simple problem modifications could solve. The teachers were mixed on how GenAI-based personalization should be implemented, pointing to the need for further research. This study shows the importance of involving teachers as co-designers as GenAI continues to unfold on the international educational stage.

Authors

• Candace A. Walkington, Southern Methodist University

• Katie Bainbridge, Rice University