GENERATIVE ARTIFICIAL INTELLIGENCE IN MATHEMATICS
EDUCATION: A SYSTEMATIC REVIEW OF DATA-DRIVEN
APPLICATIONS.
LEARNING THEORIES.
AND IMPLICATIONS
FOR SUSTAINABLE DEVELOPMENT GOAL
Khoirul Anwar1.
Adelia Sherlyna2.
Alyssa Marfinda Salsanifa3.
Havi Ayuning Tyas4.
Syahda Eka Prayudistyan5.
Genta Aldi Saputra6.
Iwan Maulana7
1,2,3,4,5,6,7
Program Studi Teknologi Pendidikan.
Universitas Negeri Surabaya.
Indonesia Email: 25112104036@mhs.
Diajukan: Maret 2026 Riwayat Artikel:
Diterima: Maret 2026 Diterbitkan: April 2026 Abstract The rapid advancement of generative artificial intelligence (AI) has introduced transformative opportunities in mathematics education, yet its implications for pedagogical practices, learning theories, and sustainable development remain underexplored.
This systematic review examines the intersection of generative AI and mathematics education, focusing on data-driven applications, theoretical frameworks, and their alignment with Sustainable Development Goal 4 (SDG .
, which advocates for inclusive and equitable quality education.
We synthesize existing research to identify key trends, challenges, and opportunities across multiple dimensions, including higher education.
STEM disciplines, adaptive learning, and ethical considerations.
analyzing diverse scholarly works, we uncover how gesnerative AI supports personalized learning, enhances problem-solving skills, and fosters engagement while addressing disparities in educational access.
The review highlights the role of generative AI in promoting active learning through interactive tools, yet it also reveals concerns regarding algorithmic bias, data privacy, and the need for teacher preparedness.
Our findings suggest that while generative AI holds significant potential to democratize mathematics education, its responsible integration requires robust pedagogical strategies and policy frameworks.
The study contributes to ongoing discussions on AI-driven educational innovation by offering evidence-based insights for researchers, educators, and policymakers aiming to harness generative AI for sustainable educational development.
Keyword : Generative AI.
Mathematics Education.
SDG 4.
Personalized Learning.
Systematic Review
INTRODUCTION
The integration of artificial intelligence (AI) into education has ushered in a new era of pedagogical innovation, with generative AI emerging as a particularly transformative force.
Unlike traditional AI systems that rely on predefined rules or static datasets, generative AI models, such as large language models (LLM.
and diffusionbased architectures, can produce novel content, simulate human-like interactions, and AL JABAR: Jurnal Pendidikan dan Pembelajaran Matematika Volume 5 Nomor 2.
April 2026 adapt to diverse learning contexts (Alasadi & Baiz, 2.
This capability has profound implications for mathematics education, a discipline often characterized by abstract concepts, procedural complexity, and varying student engagement levels.
Generative AI offers tools for personalized problem generation, real-time feedback, and interactive tutoring, thereby addressing long-standing challenges in mathematics instruction (Mohamed et al.
, 2.
Mathematics education is a cornerstone of STEM literacy and a critical enabler of Sustainable Development Goal 4 (SDG .
, which seeks to Auensure inclusive and equitable quality education and promote lifelong learning opportunities for allAy (Unterhalter, 2.
Despite its importance, mathematics remains a barrier for many learners due to factors such as instructional rigidity, limited access to qualified teachers, and socio-economic Generative AI presents an opportunity to mitigate these challenges by democratizing access to high-quality, adaptive learning resources.
For instance.
AI-driven platforms can tailor problems to individual proficiency levels, provide multilingual support, and simulate one-on-one tutoring in under-resourced settings (Bi, 2.
However, the rapid adoption of generative AI in education has outpaced rigorous empirical evaluation, leaving critical gaps in understanding its long-term pedagogical and ethical implications.
While some studies highlight its potential to enhance engagement and conceptual understanding (Gjermeni & Prodani, 2.
, others caution against overreliance on AI-generated content, which may inadvertently reinforce superficial learning or algorithmic biases (Baker & Hawn, 2.
Moreover, the theoretical foundations for integrating generative AI into mathematics education remain Existing learning theories, such as constructivism and cognitive load theory, were not designed to account for AI-mediated interactions, necessitating new frameworks to guide effective implementation (Gibson et al.
, 2.
The motivation for this systematic review stems from the need to consolidate fragmented research on generative AI in mathematics education and assess its alignment with SDG 4.
By synthesizing empirical findings, theoretical perspectives, and ethical considerations, we aim to provide a comprehensive overview of how generative AI can support or hinder sustainable educational development.
This review is significant for multiple stakeholders: educators can leverage evidence-based insights to design AIenhanced curricula, policymakers can identify regulatory priorities, and researchers can pinpoint underexplored areas for future inquiry.
The remainder of this paper is organized as follows: Section 2 outlines the methodology for literature selection and analysis.
Section 3 presents the results, structured into eight thematic subsections that explore research trends, applications in higher education and STEM, learning theories, user perceptions, ethical concerns, and adaptive learning.
Section 4 discusses the synthesized findings, and Section 5 concludes with forward-looking recommendations.
METODS
Review Protocol AL JABAR: Jurnal Pendidikan dan Pembelajaran Matematika Volume 5 Nomor 2.
April 2026 This systematic review adheres to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyse.
guidelines (Page et al.
, 2.
to ensure methodological rigor and transparency.
The literature search was conducted across nine databases and search engines, selected for their relevance to education, artificial intelligence, and interdisciplinary research.
Web of Science and Scopus were prioritized due to their extensive coverage of high-impact journals and conference proceedings in STEM education and AI.
PubMed was included to capture studies at the intersection of cognitive science and technology-enhanced learning.
Ie Xplore and ACM Digital Library provided access to technical research on AI applications, while arXiv served as a repository for preprints in machine learning and education.
SpringerLink and ScienceDirect were chosen for their robust collections of peer-reviewed educational Google Scholar was used as a supplementary tool to identify grey literature and ensure comprehensive coverage.
The search strings combined keywords related to generative AI (AuGenerative Artificial IntelligenceAy OR AuGAIA.
, mathematics education (AuMathematics EducationAy OR AuMath EducationA.
, data-driven approaches (AuData-Driven ApplicationsAy OR AuDataDriven ApproachesA.
, learning theories (AuLearning TheoriesA.
, and sustainability goals (AuSustainable Development Goal 4Ay OR AuSDG 4A.
Filters excluded review articles, surveys, and meta-analyses to focus on primary research, and the publication window was set from 2016 to the present to capture recent advancements.
Research Dimensions The analysis is structured around eight research dimensions that collectively address the multifaceted role of generative AI in mathematics education.
Generative AI in Higher Education for Sustainable Development examines how AI tools support tertiary-level learning and institutional goals aligned with SDG 4.
Generative AI in Mathematics Education focuses on subject-specific applications, such as problem-solving aids and concept visualization.
The STEM Education dimension explores crossdisciplinary synergies, particularly in physics and engineering contexts where mathematical proficiency is critical.
Generative AI and Learning & Teaching investigates pedagogical strategies, including flipped classrooms and scaffolded feedback.
User Perception and Acceptance evaluates stakeholder attitudes, while Ethical and Responsible Use highlights challenges like bias mitigation and academic integrity.
Finally.
Generative AI in Distance and Adaptive Learning assesses scalability and inclusivity in nontraditional settings.
Inclusion and Exclusion Criteria Studies were included if they: .
empirically evaluated generative AI applications in mathematics or STEM education, .
explicitly addressed at least one research dimension, .
were peer-reviewed, and .
published in English between 2016Ae2023.
Exclusion criteria removed studies lacking methodological details, those focused solely on non-generative AI .
, rule-based tutor.
, and opinion pieces without empirical data.
The timeframe ensured relevance to contemporary AI advancements, while the language restriction mitigated translation biases.
Study Selection Process The initial search yielded 771 records, reduced to 597 after deduplication and the removal of three irrelevant entries.
Title and abstract screening excluded 391 records.
AL JABAR: Jurnal Pendidikan dan Pembelajaran Matematika Volume 5 Nomor 2.
April 2026 primarily due to mismatches with the research dimensions .
AI applications in noneducational domain.
Full-text review of 122 articles led to the exclusion of 58 ineligible studies, with common reasons being insufficient focus on generative AI .
or absence of mathematics/STEM education outcomes .
The final corpus comprised 64 studies.
As shown in Figure 1, the PRISMA flowchart illustrates this attrition process.
key limitation is the potential omission of non-English studies, which may introduce geographic bias.
Additionally, the rapid evolution of generative AI means some cuttingedge applications in preprint repositories might lack peer-reviewed validation.
Figure 1.
PRISMA flowchart of the study selection process The quality assessment prioritized studies with clear research questions, robust sample sizes, and reproducible methodologies.
For example, articles employing randomized controlled trials (RCT.
or longitudinal designs were weighted more heavily than small-scale case studies.
This approach ensured the synthesis reflected evidencebased practices while acknowledging exploratory work in emerging areas.
AL JABAR: Jurnal Pendidikan dan Pembelajaran Matematika Volume 5 Nomor 2.
April 2026 Results And Discussion Research Trends Figure 2.
Research trends in generative artificial intelligence applications for mathematics education and sustainable development The analysis of publication patterns reveals a striking surge in research interest toward generative AI in mathematics education, particularly between 2023 and 2025.
While only nine studies were identified for 2023, this number more than tripled to 29 publications in 2024, followed by 23 in 2025.
This exponential growth trajectory underscores the fieldAos responsiveness to technological advancements, likely catalyzed by the widespread adoption of large language models like ChatGPT in educational The sharp decline to a single publication in 2026 should be interpreted cautiously, as this may reflect incomplete indexing of recent works rather than diminished scholarly attention.
Thematic distribution across the eight research dimensions shows uneven but complementary foci.
Generative AI in Mathematics Education dominates the corpus, accounting for 16 studies .
% of total publication.
, with nearly equal representation between 2024 .
and 2025 .
This concentration aligns with mathematics educationAos unique challenges, where generative AIAos capacity for symbolic reasoning and step-bystep problem generation offers distinct advantages over other disciplines.
The Learning & Teaching dimension follows closely, spanning 26 studies across 2023Ae2025, indicating strong interest in pedagogical integration strategies.
Notably, this theme peaked in 2024 .
before halving in 2025, suggesting initial enthusiasm may be giving way to more specialized investigations.
Emerging themes exhibit contrasting trajectories.
Ethical considerations, though modest in volume .
, maintain steady presence, reflecting growing recognition of risks such as algorithmic bias and academic dishonesty.
Conversely.
STEM Education applications remain sparse .
, revealing an opportunity for cross-disciplinary AL JABAR: Jurnal Pendidikan dan Pembelajaran Matematika Volume 5 Nomor 2.
April 2026 research that connects mathematical learning with broader scientific contexts.
The temporal clustering of Distance and Adaptive Learning studies in 2025 .
coincides with global shifts toward hybrid education models post-pandemic, highlighting how societal changes drive research priorities.
Methodologically, early studies .
3Ae2.
predominantly employ qualitative approaches, including case studies of AI tool implementation and interviews with Later works .
5 onwar.
increasingly adopt mixed-methods designs, combining learning analytics from AI platforms with psychometric assessments of student outcomes.
This evolution suggests the field is maturing from exploratory investigations toward rigorous efficacy testing.
Geographic analysis shows disproportionate representation from North America and Europe .
% of studie.
, raising questions about the generalizability of findings to Global South contexts where SDG 4 challenges are most acute.
The trends collectively paint a picture of a field in rapid transition, where technological possibilities outpace empirical validation.
While the proliferation of studies demonstrates generative AIAos perceived value for mathematics education, the concentration on higher-income regions and limited longitudinal data underscore the need for more inclusive, sustained research efforts.
Generative AI in Higher Education for Sustainable Development The integration of generative artificial intelligence (GAI) in higher education has emerged as a pivotal strategy for advancing Sustainable Development Goal 4 (SDG .
, which emphasizes inclusive and equitable quality education.
This subsection examines how GAI applications address systemic challenges in tertiary education while fostering sustainable learning ecosystems.
The included studies reveal three dominant paradigms:
overcoming institutional barriers through adaptive learning technologies, .
personalizing education to bridge equity gaps, and .
redefining pedagogical frameworks for AI-augmented classrooms.
A critical contribution of GAI lies in its capacity to democratize access to advanced mathematical instruction.
As demonstrated in (Pachava et al.
, 2.
AI-driven platforms enable real-time customization of learning materials, adapting problem difficulty and explanatory depth to individual student needs.
This aligns with SDG 4Aos 3 on ensuring equal access to affordable technical education.
For instance, generative models can simulate one-on-one tutoring for underrepresented groups in STEM, effectively compensating for regional teacher shortages (AlSagri & Sohail, 2.
However, (Jogezai et al.
, 2.
cautions that such benefits presuppose reliable digital infrastructure a requirement still unmet in many developing economies, potentially exacerbating existing educational disparities.
The transformative potential of GAI extends beyond accessibility to curricular Table 1 synthesizes key findings from the reviewed studies, categorizing them by their primary contributions to sustainable higher education.
AL JABAR: Jurnal Pendidikan dan Pembelajaran Matematika Volume 5 Nomor 2.
April 2026 Table 1.
Taxonomy of GAI Applications in Higher Education for SDG 4 Focus Key Themes Sources Area Challenge Institutional resistance, (Jogezai et al.
, 2.
& infrastructure requirements, and Opportun workforce upskilling needs SDG 4 Personalized learning pathways, (Pachava et al.
, 2.
, (AlSagri & Advance equity-focused interventions, and Sohail, 2.
multilingual support Theoretic Connectivism frameworks for AI- (Baskara, 2.
& mediated learning and sustainability Future Direction Trends & Adoption patterns of ChatGPT and (Sahar & Munawaroh, 2.
Integratio comparative analysis of global Theoretical advancements are particularly noteworthy, with (Baskara, 2.
proposing connectivism as a foundational theory for GAI-integrated education.
This perspective posits that learning in AI-augmented environments occurs through dynamic networks of human and machine interactions, challenging traditional cognitivist Meanwhile, (Sahar & Munawaroh, 2.
identifies a paradoxical trend:
while 78% of surveyed universities in North America have pilot GAI programs, only 12% have established ethical guidelines for their useAia gap that undermines the sustainability of these initiatives.
Ethical considerations permeate all application domains.
The ability of GAI to generate authentic-looking mathematical proofs and solutions raises fundamental questions about academic integrity and assessment redesign (Pachava et al.
, 2.
Furthermore, as noted in (AlSagri & Sohail, 2.
, the environmental costs of training large language models may conflict with SDG 4Aos emphasis on sustainable infrastructure.
These tensions underscore the need for holistic implementation frameworks that balance pedagogical innovation with planetary boundaries.
The reviewed studies collectively demonstrate that while GAI can accelerate progress toward SDG 4 in higher education, its sustainability hinges on addressing three interdependent factors: equitable access to technology, pedagogical adaptation by educators, and robust governance mechanisms.
Future research must prioritize longitudinal studies across diverse socioeconomic contexts to validate the long-term efficacy of these interventions.
Generative AI in Mathematics Education: Pedagogical Innovations and Challenges The integration of generative AI into mathematics education has catalyzed a paradigm shift in instructional design, student engagement, and assessment This subsection synthesizes empirical evidence from 16 studies that explore how generative AI toolsAiranging from large language models (LLM.
to AL JABAR: Jurnal Pendidikan dan Pembelajaran Matematika Volume 5 Nomor 2.
April 2026 multimodal storytelling systemsAiare reshaping the teaching and learning of mathematical concepts.
The analysis reveals three dominant themes: .
the augmentation of problem-solving and reasoning skills, .
the personalization of learning experiences, and .
the emerging challenges in teacher preparedness and curricular adaptation.
A prominent application of generative AI in mathematics education lies in its capacity to scaffold complex problem-solving processes.
Studies such as (Wardat, 2.
demonstrate how ChatGPT can generate step-by-step solutions to algebraic equations while providing adaptive feedback based on student responses.
This aligns with cognitivist theories of learning by externalizing mental models and making abstract concepts tangible.
However, (Bastani et al.
, 2.
presents counterevidence from high school classrooms, where unfettered access to GPT-4 led to a 22% decline in procedural fluency, suggesting that overreliance on AI-generated solutions may undermine foundational skill development.
The tension between these findings underscores the need for carefully structured AI interventions that balance conceptual understanding with procedural practice.
The personalization capabilities of generative AI emerge as a consistent strength across multiple studies.
Research by (H.
Li et al.
, 2.
illustrates how AI-generated mathematical stories can adapt narrative complexity and problem types based on realtime assessments of student proficiency.
This multimodal approachAicombining textual, visual, and symbolic representationsAiaddresses diverse learning styles while maintaining alignment with curricular standards.
Similarly, (M.
Li, 2.
documents how primary school teachers in under-resourced settings used generative AI to create culturally relevant word problems, resulting in a 31% increase in student engagement compared to traditional textbook exercises.
These applications resonate with Vygotskian principles of zone of proximal development, where AI tools act as dynamic scaffolds that adjust to individual learning trajectories.
Table 2.
Taxonomy of Generative AI Applications in Mathematics Education Applicatio n Type Intelligent Tutoring Pedagogical Function Step-by-step problem solving Adaptive Storytellin Contextualized Proof Generatio Teacher Support Automated theorem proving TPACK Key Findings Sources Improves conceptual understanding (Wardat, but risks procedural skill atrophy (Bastani et , 2.
Enhances through (H.
Li et culturally responsive narratives , 2.
, (M.
Li.
Effective for advanced learners but (Dilling & requires validation of logical rigor Herrmann , 2.
Increases pre-service teachersAo (Segal & problem-posing creativity Biton, (Biton & AL JABAR: Jurnal Pendidikan dan Pembelajaran Matematika Volume 5 Nomor 2.
April 2026 Applicatio n Type Pedagogical Function Key Findings Sources Segal.
The integration of generative AI into mathematics education necessitates significant shifts in teacher roles and professional development.
Studies focusing on teacher perceptions reveal both enthusiasm and apprehension: while 68% of surveyed educators in (Alsharidah & Alkramiti, 2.
acknowledged AIAos potential to reduce workload through automated assessment, 53% expressed concerns about maintaining academic rigor when students use AI tools.
This dichotomy is particularly evident in (Y.
Wang et al.
, 2.
, where mathematics teachersAo acceptance of generative AI correlated strongly with their technological self-efficacy .
= 0.
72, p < 0.
, highlighting the importance of targeted teacher training programs.
Cultural and contextual factors significantly mediate the effectiveness of AI Research by (Payadnya et al.
, 2.
in Southeast Asian classrooms demonstrates that generative AI tools achieved higher adoption rates when aligned with local pedagogical traditions for instance, by incorporating collaborative problem-solving features that reflect collectivist learning values.
Conversely, (Engelbrecht & Borba, 2.
warns against the uncritical transfer of Western-centric AI models to diverse educational contexts, noting that language models trained primarily on English datasets often struggle with mathematical terminology in other languages, potentially exacerbating educational The ethical dimension of generative AI in mathematics education surfaces repeatedly across studies.
(Opesemowo & Ndlovu, 2.
identifies three key concerns:
the Aublack boxAy nature of AI-generated solutions that may obscure mathematical reasoning, the potential for algorithmic bias in adaptive learning systems, and the environmental costs of deploying large-scale AI models in resource-constrained settings.
These challenges are compounded by the rapid evolution of generative AI technologies, which often outpace the development of appropriate pedagogical frameworks and policy Emerging research directions suggest promising avenues for future investigation.
Several studies ((Cosentino et al.
, 2.
, (Drijvers & Sinclair, 2.
) explore the integration of embodied learning theories with generative AI, using multimodal data .
gesture recognition, eye trackin.
to create immersive mathematical learning experiences.
Others ((Umoh, 2.
, (Tashtoush et al.
, 2.
) examine the longitudinal impacts of AIaugmented mathematics instruction on STEM career pathways, particularly for female and minority students who have historically been underrepresented in quantitative fields.
These developments point toward a more holistic understanding of generative AIAos role in mathematics educationAione that transcends technical functionality to address broader questions of equity, epistemology, and sustainable pedagogical practice.
Generative AI in STEM Education: Cross-Disciplinary Applications and Teacher Preparedness The application of generative artificial intelligence in STEM education represents a critical intersection between technological innovation and pedagogical transformation.
AL JABAR: Jurnal Pendidikan dan Pembelajaran Matematika Volume 5 Nomor 2.
April 2026 This subsection examines how generative AI tools are being integrated across science, technology, engineering, and mathematics disciplines, with particular attention to teacher perspectives and sustainable development applications.
The analysis draws from two key studies that illuminate both the potential and challenges of implementing these technologies in K-12 and higher education settings.
A central theme emerging from the literature is the role of teacher perceptions in shaping the adoption of generative AI in STEM classrooms.
The study by (Darayseh & Mersin, 2.
investigates science and mathematics teachersAo experiences with AI integration, revealing a spectrum of attitudes ranging from enthusiastic adoption to cautious skepticism.
Teachers who successfully incorporated generative AI into their instruction reported using it to create customized problem sets, simulate scientific experiments, and provide real-time feedback on student work.
However, significant barriers were identified, including limited professional development opportunities and concerns about maintaining academic integrity when students use AI-generated content.
These findings align with broader research on technology integration in education, which emphasizes the importance of teacher self-efficacy and institutional support in determining the success of innovative tools.
The potential of generative AI to address sustainability challenges through STEM education is particularly noteworthy.
As highlighted in (Cheah & Kim, 2.
AI tools can facilitate project-based learning focused on environmental issues by enabling students to analyze complex datasets, model climate change scenarios, and develop solutions for sustainable development.
This application directly supports Sustainable Development Goal 4Aos emphasis on education for sustainability while also fostering critical STEM competencies such as data literacy and systems thinking.
The study found that teachers who implemented these AI-enhanced projects reported increased student engagement with sustainability topics, particularly when the AI tools allowed for visualization of environmental impacts at local and global scales.
Table 3.
Framework for Generative AI Integration in STEM Education Integration Dimension Curriculum Alignment Key Characteristics AI-generated content mapped to STEM standards Sustainability Focus Teacher Support Implementation Challenges Ensuring accuracy of AI outputs for specialized Sources (Darayse Mersin.
Climate modeling Access to quality training (Cheah and environmental datasets & Kim, data analysis Professional Time constraints and (Darayse development for AI competing priorities tool integration Mersin, (Cheah & Kim.
AL JABAR: Jurnal Pendidikan dan Pembelajaran Matematika Volume 5 Nomor 2.
April 2026 Integration Dimension Ethical Considerations Key Implementation Sources Characteristics Challenges Addressing bias in Developing student (Cheah AI-generated STEM awareness AI & Kim.
The studies collectively underscore the importance of developing comprehensive support structures for STEM educators working with generative AI.
Both (Darayseh & Mersin, 2.
and (Cheah & Kim, 2.
emphasize that successful implementation requires more than just access to technologyAiit demands ongoing professional learning communities where teachers can share best practices, troubleshoot challenges, and collaboratively develop AI-enhanced lesson plans.
This need is particularly acute in interdisciplinary STEM contexts, where teachers must navigate not only the technical aspects of AI tools but also their pedagogical implications across multiple subject areas.
An emerging area of concern is the equitable distribution of AI resources across different educational contexts.
While (Cheah & Kim, 2.
demonstrates the potential of generative AI to enhance sustainability education, it also notes that schools in underresourced areas often lack the infrastructure necessary to support these applications.
This disparity raises important questions about how to ensure that the benefits of AI in STEM education are accessible to all students, regardless of their socioeconomic background or geographic location.
The studies suggest that addressing these equity issues will require coordinated efforts among policymakers, technology developers, and educational The research also highlights the evolving nature of STEM pedagogy in the age of generative AI.
Traditional approaches to teaching science and mathematics often emphasize procedural knowledge and standardized testing, but AI tools enable more open-ended, inquiry-based learning experiences.
This shift aligns with contemporary theories of STEM education that prioritize creativity, problem-solving, and real-world However, as both studies caution, realizing this potential depends on careful instructional design that leverages AIAos strengths while mitigating its limitationsAia balance that requires both technological expertise and deep pedagogical knowledge.
Generative AI and Learning & Teaching: Pedagogical Transformations The integration of generative artificial intelligence (AI) into learning and teaching processes has catalyzed significant pedagogical innovations while simultaneously challenging traditional educational paradigms.
This subsection examines how generative AI tools are reshaping instructional strategies, student learning experiences, and theoretical frameworks across diverse educational contexts.
The analysis reveals three critical dimensions of this transformation: .
the redefinition of teacher and learner roles, .
the emergence of new instructional models, and .
the adaptation of learning theories to accommodate AI-mediated education.
Generative AI has fundamentally altered the dynamics between teachers and learners, creating a more interactive and personalized educational environment.
Studies such as (Lypez et al.
, 2.
demonstrate how AI-powered tools in higher education can function as Aupedagogical partners,Ay assisting instructors in developing customized learning materials for disciplines like business mathematics while enabling students to AL JABAR: Jurnal Pendidikan dan Pembelajaran Matematika Volume 5 Nomor 2.
April 2026 engage in self-directed problem-solving.
This dual capacity challenges traditional teacher-centered models, fostering a more distributed form of pedagogical responsibility.
However, research by (Cabellos et al.
, 2.
reveals significant variation in teacher acceptance, with only 42% of surveyed faculty expressing confidence in adapting their teaching methods to incorporate generative AI effectively.
The tension between AIAos potential to enhance instruction and educatorsAo readiness to harness this potential emerges as a recurring theme across multiple studies.
The pedagogical applications of generative AI extend across various instructional models, as detailed in Table 4.
Conversational AI systems like ChatGPT have been particularly transformative, enabling new forms of dialogic learning that promote deeper engagement with complex concepts.
Research by (Matthew et al.
, 2.
develops a framework for implementing ChatGPT in educational settings, emphasizing its capacity to simulate Socratic questioning techniques and provide instant feedback.
Similarly, (Urban et al.
, 2.
demonstrates how generative AI can enhance creative problemsolving in university students, with experimental data showing a 28% improvement in solution originality when AI tools are used as cognitive scaffolds rather than answer These findings suggest that the most effective applications of generative AI in education occur when the technology complements rather than replaces human cognitive processes.
Table 4.
Generative AI Applications in Learning and Teaching Application Domain Conversation al AI Pedagogical Impact Key Challenges Sources Enables dialogic learning Requires careful prompt (Matthew et and instant feedback , 2.
(Szaby Szoke.
Creative Enhances solution Risk of over-reliance on (Urban et al.
Problemoriginality and flexibility AI-generated ideas Solving (Song et al.
Personalized Adapts to Data and (Lytvynova Learning learning algorithmic bias concerns et al.
, 2.
(Wei et al.
Teacher Supports AI-augmented Limited institutional (Lee & Zhai.
Professional lesson planning training opportunities Development (Alexandro wicz, 2.
The theoretical implications of generative AI for learning and teaching are profound, necessitating revisions to established educational frameworks.
As noted in (Koh & Doroudi, 2.
, the advent of generative AI requires educators to Auregenerate older learning theoriesAy to account for AI-mediated cognitive processes.
Constructivist approaches, for instance, must now consider how learners construct knowledge not only through human interactions but also through engagements with AI systems.
Similarly.
AL JABAR: Jurnal Pendidikan dan Pembelajaran Matematika Volume 5 Nomor 2.
April 2026 (Gibson et al.
, 2.
explores how mathematical learning theories can be adapted to incorporate AIAos capacity for topological data analysis and generative representations.
These theoretical adaptations are particularly crucial for mathematics education, where AIAos ability to visualize abstract concepts and generate infinite practice problems challenges traditional notions of skill acquisition and mastery.
Disciplinary differences in generative AI adoption present another layer of Research by (Qu et al.
, 2.
reveals significant variation in how undergraduate students engage with AI tools across fields, with mathematics and physics students demonstrating more sophisticated usage patterns compared to their humanities This divergence suggests that the effectiveness of generative AI in learning and teaching may be contingent on the epistemological structure of the discipline, with rule-based domains like mathematics being particularly amenable to AI augmentation.
The ethical dimension of generative AI in education remains a persistent concern across studies.
(Kaplan-Rakowski et al.
, 2.
identifies three primary ethical challenges in AI-mediated teaching: the potential for diminishing human interaction in learning processes, the risk of propagating biases present in training data, and the environmental costs associated with large-scale AI deployments.
These concerns are compounded by the rapid pace of technological advancement, which often outstrips the development of appropriate pedagogical and ethical guidelines.
Emerging research directions point toward increasingly sophisticated integrations of generative AI in educational settings.
Studies such as (Wei et al.
, 2.
explore the use of multiple AI pedagogical agents in augmented reality environments, creating immersive learning experiences that blend physical and digital interactions.
Others, like (Kostopoulos et al.
, 2.
, investigate Auagentic AIAy systems that can autonomously adjust teaching strategies based on real-time analysis of student performance data.
These developments suggest a future where generative AI becomes not just a tool for learning, but an active participant in the educational processAia transformation that will require ongoing critical examination of its pedagogical, ethical, and social implications.
The studies collectively demonstrate that while generative AI holds tremendous potential to enhance learning and teaching, its successful integration depends on thoughtful pedagogical design, robust teacher support systems, and continuous evaluation of both its benefits and limitations.
As the technology continues to evolve, so too must our understanding of how it can best serve educational goals while preserving the essential human elements of teaching and learning.
User Perception and Acceptance of Generative AI in Education The adoption of generative artificial intelligence (AI) in educational settings is fundamentally shaped by user perceptions and acceptance levels among students, educators, and institutional stakeholders.
This subsection synthesizes findings from seven key studies that examine the psychological, cultural, and institutional factors influencing how generative AI is received and integrated into learning environments.
The analysis reveals complex interplays between technological affordances, pedagogical beliefs, and ethical considerations that collectively determine the trajectory of AI adoption in AL JABAR: Jurnal Pendidikan dan Pembelajaran Matematika Volume 5 Nomor 2.
April 2026 A dominant theme across the reviewed studies is the application of established behavioral theories to understand AI acceptance patterns.
The study by (C.
Wang et al.
integrates the Theory of Planned Behavior with AI literacy constructs, demonstrating that studentsAo intention to use generative AI tools is strongly predicted by their attitudes ( = 0.
38, p < 0.
, subjective norms ( = 0.
29, p < 0.
, and perceived behavioral control ( = 0.
21, p < 0.
This finding suggests that acceptance is not merely a function of technological capability but is deeply embedded in social and cognitive Similarly, (Yilmaz et al.
, 2.
develops and validates a Generative AI Acceptance Scale grounded in the Unified Theory of Acceptance and Use of Technology (UTAUT), identifying performance expectancy and effort expectancy as the most significant predictors of adoption among university students.
These theoretical approaches provide robust frameworks for explaining variance in AI acceptance across different educational contexts.
Cultural and institutional contexts emerge as critical mediators of AI perception.
Research by (Adarkwah et al.
, 2.
in Ghanaian higher education institutions applies Diffusion of Innovation Theory to examine academic staff responses to ChatGPT, revealing distinct adoption patterns based on disciplinary backgrounds.
Mathematics and computer science educators showed significantly higher acceptance rates .
%) compared to humanities faculty .
%), suggesting that the perceived relevance of generative AI varies substantially across academic domains.
This study also highlights the importance of localizing AI tools, as participants emphasized the need for models trained on African educational content to ensure cultural relevance and linguistic Table 5.
Factors Influencing User Acceptance of Generative AI in Education Determinant Category Psychological Key Factors Impact Level Representative Studies self- High ( = 0.
35- (C.
Wang et al.
, 2.
(Yilmaz et al.
, 2.
Social Peer Moderate .
= (Adarkwah et al.
, 2.
(Obenza et al.
, 2.
Pedagogical Alignment Varied by (Alshammari & Alwith teaching discipline Enezi, 2.
, (Adarkwah et al.
, 2.
Ethical Concerns High for (Intelligence in Higher EducationAo et al.
, 2.
(Obenza et al.
, 2.
The student perspective on generative AI reveals both enthusiasm and critical (Obenza et al.
, 2.
examines undergraduate perceptions across multiple universities, finding that 68% of respondents viewed ChatGPT as AutransformativeAy for their learning, particularly for mathematics problem-solving and writing tasks.
However.
AL JABAR: Jurnal Pendidikan dan Pembelajaran Matematika Volume 5 Nomor 2.
April 2026 the same study uncovered significant concerns about overreliance, with 41% of students reporting decreased confidence in their independent problem-solving abilities after prolonged AI use.
This paradox underscores the dual-edge nature of generative AI in educationAiwhile enhancing accessibility and efficiency, it may inadvertently undermine the development of foundational cognitive skills if not implemented thoughtfully.
Educator perceptions present a more nuanced picture, often characterized by cautious optimism tempered by practical concerns.
Research by (Alshammari & AlEnezi, 2.
on pre-service social studies teachers demonstrates how acceptance levels evolve through direct experience with AI tools.
Initial apprehension about technological complexity gave way to appreciation for AIAos capacity to generate differentiated lesson plans, though participants remained wary of potential deskilling effects on their own pedagogical creativity.
This finding aligns with (Intelligence in Higher EducationAo et al.
Aos investigation of AI misuse in academic tasks, which found that faculty concerns centered less on the technology itself and more on the need for revised assessment strategies that account for AI-assisted work.
The ethical dimension of AI acceptance emerges as a cross-cutting concern.
Multiple studies ((C.
Wang et al.
, 2.
, (Intelligence in Higher EducationAo et al.
, 2.
identify academic integrity as the foremost barrier to unreserved adoption, particularly in mathematics education where AI can generate solutions to complex problems.
However, (Yilmaz et al.
, 2.
Aos validation study reveals an interesting disconnectAiwhile educators ranked ethical concerns highly, students placed greater emphasis on practical utility, suggesting generational differences in prioritization that may shape institutional AI policies.
Emerging research directions point toward more sophisticated models of AI acceptance that account for dynamic human-AI collaboration.
The micro-macro-meso framework proposed in (Alshammari & Al-Enezi, 2.
offers a promising approach by examining acceptance at individual, institutional, and societal levels simultaneously.
This multi-layered perspective is particularly relevant for mathematics education, where AI tools must be evaluated not only for their technical capabilities but also for their alignment with long-term learning goals and societal needs.
Future studies would benefit from longitudinal designs that track how perceptions evolve as users gain more experience with increasingly advanced generative AI systems.
The reviewed studies collectively demonstrate that user acceptance of generative AI in education is neither uniform nor static, but rather a complex negotiation between technological possibilities, pedagogical values, and ethical considerations.
Successful integration strategies must therefore address not only the functional aspects of AI tools but also the human factors that ultimately determine their educational impact.
Ethical and Responsible Use of Generative AI in Education The rapid integration of generative artificial intelligence (AI) into educational settings has precipitated critical ethical debates that challenge traditional pedagogical norms and institutional policies.
This subsection examines the complex landscape of ethical considerations surrounding generative AI in education, drawing upon three pivotal studies that address issues of academic integrity, educational justice, and moral AL JABAR: Jurnal Pendidikan dan Pembelajaran Matematika Volume 5 Nomor 2.
April 2026 justification for restrictive policies.
The analysis reveals tensions between innovation and regulation, access and equity, as well as human creativity and algorithmic output.
A fundamental ethical concern centers on the redefinition of authorship and creativity in AI-augmented learning environments.
As articulated in (Rahimi & Azadmanesh, 2.
, the emergence of generative AI has profoundly disrupted conventional notions of intellectual property and originality in mathematics education.
The study employs deconstruction theory to analyze how AI-generated mathematical proofs and solutions complicate traditional assessments of student work, particularly when distinguishing between human-derived and machine-assisted reasoning.
This philosophical examination reveals an epistemological crisis in mathematics education, where the veracity and ownership of AI-produced content remain contested.
The research suggests that current academic integrity frameworks may be inadequate for addressing these novel challenges, necessitating revised pedagogical approaches that explicitly account for AI collaboration while preserving authentic learning outcomes.
The moral dimensions of institutional responses to generative AI adoption present another critical ethical consideration.
(Fine Licht, 2.
provides a rigorous ethical analysis of the Aubanning approachAy to generative AI in higher education, arguing that such restrictive policies can be morally justified under specific circumstances.
The study identifies three conditions where prohibitions may be ethically warranted: when AI tools perpetuate systemic inequities by advantaging students with greater technological access, when they undermine the development of essential cognitive skills through overreliance, and when they compromise assessment validity in foundational courses.
However, the research cautions against blanket bans, advocating instead for context-sensitive policies that balance innovation with educational integrity.
This nuanced perspective highlights the ethical complexity of regulating emerging technologies in learning environments, where both unrestricted access and complete prohibition may have detrimental Table 6.
Ethical Framework for Generative AI in Education Ethical Principle Academic Integrity Generative Challenge Distinguishing human AI-generated work Proposed Mitigation Sources Strategy Develop AI-aware (Rahimi & assessment rubrics Azadmanes (Fine Licht.
Educational Equitable access to AI Institutional support for (Fine Licht.
Justice disadvantaged students Pedagogical Preserving teacher Professional development (Swindell et Autonomy discretion in AI use on ethical integration , 2.
Creative Maintaining human Explicit documentation of (Rahimi & Authenticity authorship in learning AI assistance Azadmanes h, 2.
The call for comprehensive ethical frameworks emerges as a unifying theme across the reviewed studies.
(Swindell et al.
, 2.
presents a conceptual model for responsible AI integration in education, emphasizing the need for multi-stakeholder AL JABAR: Jurnal Pendidikan dan Pembelajaran Matematika Volume 5 Nomor 2.
April 2026 collaboration in developing guidelines that address both technical and philosophical The proposed framework identifies four pillars of ethical AI use: transparency in system operations, accountability for educational outcomes, fairness in access and application, and sustainability in implementation.
Particularly relevant to mathematics education is the frameworkAos emphasis on Auexplainable AIAy for mathematical problemsolving, ensuring that students and educators can interrogate the reasoning behind AIgenerated solutions rather than accepting them as opaque outputs.
The studies collectively underscore the inadequacy of current ethical paradigms to fully address the challenges posed by generative AI in education.
While (Rahimi & Azadmanesh, 2.
focuses on deconstructing traditional notions of authorship and (Fine Licht, 2.
examines the morality of restrictive policies, (Swindell et al.
, 2.
bridges these perspectives by proposing actionable principles for responsible implementation.
This tripartite analysis reveals that ethical considerations must evolve beyond simplistic binaries of permission/prohibition to engage with the nuanced ways generative AI is transforming educational practices.
Future research directions suggested by these studies include empirical investigations of AI attribution practices in student work, longitudinal studies of skill development in AI-augmented learning environments, and comparative analyses of institutional policies across cultural contexts.
The ethical imperative extends beyond immediate classroom concerns to broader societal implications.
As generative AI becomes increasingly sophisticated in mathematical reasoning and problem-solving, questions arise about its long-term impact on human mathematical cognition and the valuation of mathematical skills in the The reviewed studies suggest that educational institutions bear a responsibility not only to mitigate potential harms but also to actively shape the development of generative AI systems that align with pedagogical values and sustainable educational This proactive approach requires ongoing dialogue between educators, researchers, policymakers, and AI developers to ensure that technological advancements serve rather than subvert the fundamental purposes of education.
Generative AI in Distance and Adaptive Learning The integration of generative artificial intelligence into distance and adaptive learning environments represents a paradigm shift in how educational experiences can be personalized and scaffolded for diverse learners.
This subsection examines the transformative potential of generative AI to enhance self-regulated learning through realtime analytics and adaptive scaffolding, while also exploring its role in making distance education more interactive and personalized through simulated learning experiences.
A critical advancement in this domain is demonstrated by (T.
Li et al.
, 2.
which presents a framework for converting real-time learning analytics into adaptive scaffolds using generative AI.
The study reveals how learner trace data can be dynamically filtered to provide targeted support, with the AI system withholding scaffolds when learners demonstrate sufficient competence.
This approach aligns with VygotskyAos zone of proximal development while addressing the challenge of overscaffolding that often plagues traditional adaptive learning systems.
The research highlights the delicate balance required in scaffold timing and content generation, where AL JABAR: Jurnal Pendidikan dan Pembelajaran Matematika Volume 5 Nomor 2.
April 2026 premature or excessive AI intervention can inadvertently hinder the development of independent problem-solving skills.
The personalization capabilities of generative AI in distance education are further explored in (Christadoss & Panda, 2.
, which investigates simulated learning These AI-generated simulations create immersive, context-rich scenarios that adapt to individual learner progress, effectively bridging the transactional distance that often characterizes remote education.
The study identifies three key benefits:
increased learner engagement through personalized narrative structures, improved conceptual understanding via adaptive problem generation, and enhanced metacognitive awareness through AI-facilitated reflection prompts.
However, the research also cautions against potential over-reliance on simulated environments, emphasizing the need to maintain authentic human interaction components in distance learning designs.
Table 7.
Applications of Generative AI in Distance and Adaptive Learning Applicatio n Domain Adaptive Scaffoldin Simulated Learning Key Functionality Real-time analytics-driven AI-generated Pedagogical Benefits Promotes selfregulated learning Implementatio Sources n Challenges Determining (T.
Li et optimal scaffold al.
, 2.
Enhances Balancing (Christad and simulation with oss Panda.
Personaliz AI-curated Addresses individual Ensuring (Laak & adaptive learning learning needs Aru.
Learning The study by (Laak & Aru, 2.
on AI and personalized learning complements these findings by examining the broader implications of generative AI for adaptive education systems.
The research identifies a critical tension between personalization and standardization, where AI systems must navigate the competing demands of individualized learning trajectories and institutional assessment requirements.
This challenge is particularly acute in mathematics education, where generative AIAos capacity to create infinite practice variations must be carefully coordinated with curriculum standards and learning objectives.
The ethical dimensions of generative AI in distance learning emerge as a recurring concern across the reviewed studies.
Issues of data privacy, algorithmic transparency, and equitable access are amplified in remote learning contexts where students may have varying levels of technological infrastructure and support.
The studies collectively suggest that while generative AI holds tremendous potential to democratize access to quality education through adaptive distance learning, its implementation must be guided by robust ethical frameworks that prioritize learner autonomy and equitable outcomes.
Emerging research directions point toward increasingly sophisticated integrations of generative AI with other educational technologies.
The combination of adaptive learning systems with virtual reality environments, for instance, could create immersive mathematics learning experiences that respond dynamically to learner actions and cognitive states.
Similarly, the integration of affective computing with generative AI AL JABAR: Jurnal Pendidikan dan Pembelajaran Matematika Volume 5 Nomor 2.
April 2026 could enable systems to adapt not only to cognitive needs but also to emotional states, potentially reducing the isolation often associated with distance learning.
These advancements, while promising, will require continued research to ensure they enhance rather than replace the human elements of education that remain essential for deep learning and motivation.
The reviewed studies collectively demonstrate that generative AI is reshaping the landscape of distance and adaptive learning by enabling unprecedented levels of personalization and interactivity.
However, the successful implementation of these technologies depends on careful pedagogical design, ongoing evaluation of learning outcomes, and thoughtful consideration of the ethical implications inherent in AImediated education.
As the field continues to evolve, it will be critical to maintain a learner-centered approach that harnesses the power of generative AI while preserving the essential human dimensions of teaching and learning.
The synthesis of findings across the reviewed studies reveals a complex landscape where generative artificial intelligence (AI) simultaneously disrupts and enhances mathematics education.
Taken together, the research demonstrates that generative AI applications consistently improve accessibility and personalization in learning, yet they also introduce novel challenges that require careful pedagogical and ethical consideration.
The patterns emerging across studies suggest that while AI tools like ChatGPT and adaptive tutoring systems show promise in scaffolding mathematical problem-solving (Wardat, 2.
, their effectiveness is mediated by factors such as implementation context, teacher preparedness, and student self-regulation (Bastani et al.
, 2.
The theoretical implications of these findings are profound, necessitating revisions to established learning frameworks.
Constructivist and sociocultural theories must now account for AI as an active participant in the learning process, not merely a For instance, the connectivist perspective proposed by (Baskara, 2.
offers a viable model for understanding how knowledge construction occurs through human-AI interactions, particularly in distance learning environments where generative AI provides adaptive scaffolding (T.
Li et al.
, 2.
However, this theoretical expansion raises fundamental questions about the nature of mathematical understanding when learners collaborate with systems capable of generating proofs and solutions (Rahimi & Azadmanesh, 2.
The tension between procedural fluency and conceptual understandingAilong debated in mathematics educationAitakes on new dimensions when AI can perform procedural tasks effortlessly, potentially altering the cognitive goals of mathematics instruction.
Practically, the findings underscore the need for institutional support systems that prepare educators to integrate generative AI effectively.
The consistent theme of teacher apprehension across studies (Alsharidah & Alkramiti, 2.
, coupled with evidence of enhanced student outcomes when AI is implemented thoughtfully (H.
Li et al.
, 2.
suggests that professional development programs should focus on developing technological pedagogical content knowledge (TPACK) specific to AI tools.
Moreover, the disparities in AI access and adoption between high- and low-resource settings (Jogezai et al.
, 2.
highlight the urgency of policy initiatives that ensure equitable distribution of these technologies, aligning with SDG 4Aos emphasis on inclusive education.
AL JABAR: Jurnal Pendidikan dan Pembelajaran Matematika Volume 5 Nomor 2.
April 2026 Several methodological limitations in the reviewed literature temper the generalizability of these findings.
The predominance of small-scale, short-duration studies in high-income contexts (C.
Wang et al.
, 2.
limits insights into long-term impacts and cross-cultural applicability.
Publication bias toward positive outcomes may obscure instances where generative AI implementations failed or produced neutral Furthermore, the rapid evolution of generative AI technologies means that studies conducted even one year apart may examine substantially different system capabilities, making cumulative knowledge-building challenging.
These limitations collectively suggest that while the current evidence base is promising, it remains provisional.
Future research should prioritize longitudinal studies that track the sustained effects of generative AI on mathematical skill development and STEM career pathways.
There is a critical need for investigations in underrepresented educational contexts, particularly in Global South regions where infrastructure challenges may differentially impact AIAos benefits (Adarkwah et al.
, 2.
The development of standardized assessment frameworks that can distinguish between human and AI-generated problemsolving processes would address pressing academic integrity concerns while enabling more nuanced studies of learning outcomes.
Additionally, interdisciplinary collaborations between mathematics educators.
AI ethicists, and cognitive scientists could yield innovative frameworks for evaluating how generative AI influences mathematical cognition at neurological and behavioral levels.
The ethical dimensions of generative AI in mathematics education demand continued scholarly attention.
While current studies identify key concerns such as algorithmic bias and environmental costs (Opesemowo & Ndlovu, 2.
, there remains a paucity of research on culturally responsive AI design and the long-term societal implications of AI-mediated mathematical reasoning.
Future work should explore how generative AI systems can be designed to promote equitable participation in mathematics while preserving the disciplineAos intellectual rigor.
The tension between innovation and tradition in mathematics pedagogy evident in debates over calculator use decades agoAi now reemerges with far greater complexity, requiring thoughtful dialogue among all educational stakeholders.
The collective evidence suggests that generative AIAos most significant contribution to mathematics education may lie in its ability to democratize access to highquality, personalized instruction.
However, realizing this potential without exacerbating existing inequities or compromising educational integrity will require coordinated efforts across research, practice, and policy domains.
As the field moves forward, maintaining a critical yet open-minded stance toward generative AIAos possibilities will be essential for harnessing its benefits while mitigating its risks in service of sustainable educational CONCLUSION This systematic review has examined the multifaceted role of generative artificial intelligence in mathematics education, addressing its data-driven applications, theoretical foundations, and implications for Sustainable Development Goal 4.
The synthesis of 64 studies reveals that generative AI holds significant potential to transform mathematics AL JABAR: Jurnal Pendidikan dan Pembelajaran Matematika Volume 5 Nomor 2.
April 2026 education through personalized learning, adaptive scaffolding, and enhanced However, its integration is not without challenges, including ethical concerns, teacher preparedness, and equitable access.
The findings underscore the need for robust pedagogical frameworks that account for AI-mediated learning while preserving the integrity of mathematical reasoning.
Theoretically, the review highlights the necessity of adapting existing learning theories to incorporate human-AI collaboration, particularly in distance and adaptive learning environments.
Practically, the evidence calls for institutional policies that balance innovation with ethical considerations, ensuring that generative AI serves as a tool for inclusive education rather than a source of further disparity.
Future research should prioritize longitudinal studies across diverse educational contexts, with particular attention to underrepresented regions.
Investigations into the cognitive and societal impacts of AI-augmented mathematics learning will be critical for developing sustainable implementation strategies.
generative AI continues to evolve, interdisciplinary collaboration among educators, researchers, and policymakers will be essential to harness its potential while addressing its limitations.
This review provides a foundation for such efforts, offering insights that can guide both practice and future inquiry in this rapidly advancing field.
REFERENCES