Generative AI Assistance for Lesson Planning to Support Positive Student Mathematics Motivation and Emotions

• In Event: Leveraging AI and Virtual Characters in Education: Insights and Impacts for Student Motivation

Abstract

Objectives and Framework
Positive motivation is not something that exists merely within students; it can be fostered through teacher actions and the classroom environment (Patrick et al., 2016). These actions have implications for how students view themselves and subjects, such as mathematics, which in turn has implications for their engagement and performance in these subjects (e.g., Fast et al., 2010; Wang et al., 2017; Yu & Singh, 2018). Although there has long been research around how teachers can support student motivation (Ames, 1992; Skinner & Belmont, 1993), creating lessons and materials that can motivate each student can be draining for teachers, who have a number of external pressures on their time. With the improved capability and accessibility of generative AI tools, such as ChatGPT, teachers have begun to turn to these tools for assistance in creating lesson content (Maiorca et al., 2024; Zhang & Tur, 2024). However, there is still much to learn about how teachers use ChatGPT for specific purposes and whether this use results in improved outcomes for students.
We report on how eight upper elementary teachers leveraged ChatGPT to improve the motivational content of their mathematics lessons by linking lessons to student interests ascertained from a student interest inventory. Teachers developed and implemented three-four lessons using ChatGPT themed around student interests (e.g., fraction addition and Minecraft; music videos and subtraction with regrouping) and taught those lessons during their normal instructional time. Using Control-Value Theory as a framework (Pekrun, 2006), we examined how ChatGPT use related to teacher actions for enhancing control and value perspectives and how these related to student perceptions of competence and value and student academic emotions. Our conceptual model is seen in Figure 1.

Methods and Materials
Researchers observed five lessons for each teacher (two or three each wherein ChatGPT was used) and rated motivationally-supportive and non-supportive behaviors using a checklist with “never,” “once or twice,” and “three+ times.” Ordinal logistic regressions, controlling for teacher so as to only examine within-teacher variability, estimated the association between ChatGPT and each behavior.
Students completed surveys of emotions and motivation before and after each lesson for 10 lessons per teacher, up to four of which included ChatGPT content. Multilevel regressions (lessons nested in students) regressed each post-lesson outcome on whether the lesson included ChatGPT content, pre-lesson rating, and teacher dummies.

Results and Scholarly Significance
Teachers used statistically significantly more non-standard examples and utility messages (relating math to both academics and life outside school), gave more specific feedback, and engaged in less lesson-unrelated chit-chat during ChatGPT lessons than non-ChatGPT lessons. In addition, student reports of boredom were statistically significantly lower and interest were statistically significantly higher in ChatGPT lessons compared to non-ChatGPT lessons. Although other emotions and motivations were not statistically significantly different between conditions, all nine outcomes were in the expected direction (Figure 2), lending support to the effectiveness of teachers’ use of ChatGPT to support student motivation with more student-relevant and engaging lesson content. Results demonstrate the potential of ChatGPT to assist teachers in creating motivationally-supportive lessons.

Authors

• Teomara Rutherford, University of Delaware

• Andrew Rodrigues, University of Delaware

• Sotheara Veng, University of Delaware

• Yiqin Cao, University of Delaware

• Santiago Duque-Baird, University of Delaware

• Ekaterina Bergwall, University of Delaware

• Rosa Julia Mykyta-Chomsky, University of Delaware

• Kristin Chisholm, University of Delaware