Exploring Fraction Comprehension and Interest in Elementary Education Through AI-Powered Personalized Learning

Synthesis: Examines AI-powered personalized learning in elementary fraction instruction through a systematic review, quantitative study (N=120), and qualitative teacher interviews. Found that AI-adaptive platforms significantly improved fraction comprehension for students with math learning difficulties compared to traditional instruction. AI personalization increased student interest and engagement. Teachers reported AI tools helped differentiate instruction but required professional development for effective integration. Adaptive Learning, Personalized Learning, K-12, Math Education, and Generative AI. Examines AI-powered personalized learning in elementary fraction instruction through a systematic review, quantitative study (N=120), and qualitative teacher interviews. Found that AI-adaptive platforms significantly improved fraction comprehension for students with math learning difficulties compared to traditional instruction. AI personalization increased student interest and engagement. Teachers reported AI tools helped differentiate instruction but required professional development for effective integration.

What this means for practice

• Instructors. Keep teacher-led instruction in the loop: in this trial the business-as-usual group showed significantly greater fraction-comprehension gains than the Mathbot group, and the time-by-condition interaction was not significant for either comprehension or interest, so AI tools must be integrated with traditional methods rather than used in isolation.
• Learners. Use AI-supported practice as a supplement to teacher feedback rather than a substitute, and ask your teacher to check the reasoning behind your fraction answers instead of trusting the tool's explanations.
• Instructors. Assess comprehension with items that require explaining reasoning: the authors caution that multiple-choice fraction items can be answered by guessing, which may inflate apparent gains.
• Administrators. Weigh licensing costs against equitable access before scaling a platform — the authors flag the cost of tools like Mathbot as a scalability barrier for under-resourced schools.

Limitations

• The final analytic sample was 22 students (13 Mathbot, 9 business-as-usual) after recruitment of 43 fell to 29 through attrition and 7 outliers were removed — below the 29 participants the power analysis had targeted.
• The intervention lasted five consecutive school days in 40-minute sessions, far short of the 9–12 weeks the authors note is recommended for evaluating special education technology, so long-term retention was not observed.
• The comparison group was fourth graders while the Mathbot group was fifth graders, so developmental differences in mathematical reasoning confound the group comparison.
• Situational interest was self-reported, with social desirability and recall bias possible, and Mathbot's black-box design prevents identifying which of its features produced any effect; the study also ran in one suburban southeastern U.S. school with a single AI tool.

Citation

Kenneth Holman (2024). Exploring Fraction Comprehension and Interest in Elementary Education Through AI-Powered Personalized Learning. PhD dissertation, University of Central Florida.