AI evidence extraction

AI in Students’ Mathematics Learning: A PRISMA-Based Systematic Review of Challenges and Solutions (2021-2025)

Yaqian Song, Mohamed Yusoff Mohd Nor, Bity Salwana Alias · International Journal of Learning Teaching and Educational Research · 2026

AI-generated evidence extraction, verified across multiple analytical personas. Not a substitute for the peer-reviewed original.

5/10

Relevance

0/4

Quality (LMQS)

Evidence

Citations

0.00

FWCI

This is an AI analysis. Read the peer-reviewed original at the publisher: https://doi.org/10.26803/ijlter.25.3.12

Methodology & findings

Study design

Systematic literature review following PRISMA framework.

Main result

The review identified that "Three primary types of AI technologies were identified: adaptive learning systems, intelligent tutoring systems, and chatbots" and that "The reviewed studies predominantly employed quasi-experimental designs with short intervention durations and small sample sizes, with most research being conducted in primary education settings." The findings further indicate that "China contributed the largest number of publications" and that "three major categories of challenges and solutions" were identified: "research-related, technical and design-related, and educational-pedagogical."

Reports effect sizes.

Research paradigm

Positivist/empiricist

Author conclusions

The authors conclude that "this study offers evidence-based insights to support researchers, developers, educators, and policymakers in aligning AI technologies with the cognitive and problem-solving demands of mathematics learning, informing more focused instructional design."

Risk of bias

Not explicitly stated in abstract. Potential risks include publication bias (studies from WoS and Scopus only), selection bias (language restrictions not mentioned but likely present), and geographic bias (China contributed largest number of publications).; Publication bias (only studies from WoS and Scopus included); Language bias (not specified if non-English studies were included); Small sample sizes in reviewed studies; Short intervention durations in reviewed studies; Quasi-experimental designs predominant in reviewed studies; Inclusion limited to quantitative studies only (potential publication bias against qualitative research); Language bias (specific languages of included studies not stated); Geographic concentration: "China contributed the largest number of publications" (potential geographic bias); Study design homogeneity: "predominantly employed quasi-experimental designs" (selection bias in included studies); Small sample sizes in primary studies increase risk of Type II error

Open questions raised

The authors identify that "despite growing interest, AI usage in mathematics learning remains in its early stages" and note "Existing studies demonstrate the limited integration of AI tools into mathematics-specific pedagogy and insufficient focus on students' mathematical problem-solving abilities, while there remains a need for the challenges associated with AI implementation to be systematically categorized."
The review addresses gaps by noting that "Existing studies demonstrate the limited integration of AI tools into mathematics-specific pedagogy and insufficient focus on students' mathematical problem-solving abilities, while there remains a need for the challenges associated with AI implementation to be systematically categorized."
Limited integration of AI tools into mathematics-specific pedagogy
Insufficient focus on students' mathematical problem-solving abilities in existing studies
Need for systematic categorization of challenges associated with AI implementation
Research predominantly in primary education settings; gaps in secondary and higher education

Data: not_statedCode: not_statedExtracted from: pdfAgreement 58%

Explore related topics

Doctoral Training