A comprehensive bibliometric analysis of 2800 AI in education papers from 1990-2024 reveals explosive growth post-2018 driven by generative AI, with significant concentration in higher education and medical fields, but concerning gaps in ethical governance frameworks and K-12 research.
Objective: This study systematically analyzed global publication patterns, keyword trends, and research priorities in AI education from 1990 to 2024 to quantify disciplinary evolution, ethical governance deficits, and global collaboration imbalances. The research aimed to move beyond descriptive bibliometrics to provide explanatory diagnostics for critical field imbalances while capturing disruptions triggered by generative AI post-2018.
Methods: The researchers conducted a comprehensive bibliometric analysis using Web of Science data, screening 2952 papers and analyzing 2800 English publications following a standardized four-stage framework (data collection, cleaning, analysis, validation). The search strategy combined education and AI terms with specific title requirements. Custom Python scripts (Pandas v2.0.3, Matplotlib v3.7.2, Seaborn v0.11.2) were developed alongside VOSviewer v1.6.20 for visualization and network analysis. The methodology included performance analysis to quantify trends and science mapping to visualize thematic structures. Data cleaning followed established protocols with duplicate removal, institutional name standardization, and manual relevance verification.
Key Findings: The analysis revealed dramatic growth patterns, with publications increasing from sporadic papers in the 1990s to 612 in 2023 and 1216 by November 2024. The US and China emerged as leading contributors (529 and 518 papers respectively), with the University of London and University of California system as core institutions. Keywords evolved significantly from "AI" and "machine learning" (2018-2020) to "ChatGPT" and "ethics" (post-2022), reflecting dual technological and ethical focuses. Among highly cited papers, 68% addressed ethical controversies, while higher education dominated with 19 of the top 50 papers, compared to only 5 focusing on K-12 education. Medical education emerged as a prominent application domain with 8 papers in the top 50. Cross-disciplinary patterns showed education studies comprising the largest research category, though papers in education journals received significantly higher citations than those in computer science journals. The study identified four dominant knowledge clusters: generative AI applications, ethics and policy frameworks, AI technology trends, and higher education applications.
Implications: The findings reveal several critical implications for AIEd development. The concentration in higher education (89% bias) indicates significant neglect of K-12 and early childhood education, potentially limiting AI's transformative impact on foundational learning. The ethics-governance gap is particularly concerning, with 78% of studies lacking actionable ethical frameworks despite widespread acknowledgment of ethical challenges. The geographic concentration in the Global North highlights structural inequalities in AI education research, potentially excluding diverse perspectives and needs. The disciplinary citation disconnect between education and computer science journals suggests problematic silos that hinder interdisciplinary collaboration. The short-term focus on immediate generative AI applications rather than long-term impacts indicates a need for sustained longitudinal research to understand lasting educational effects.
Limitations: Several important limitations affect the scope and reliability of findings. The exclusive focus on Web of Science may have excluded relevant research from other databases or conference proceedings, potentially missing innovative approaches. The geographic concentration of studies primarily from the US, China, and a few other developed countries limits cultural diversity and international applicability. The rapid pace of AI development means some findings may quickly become outdated, particularly regarding generative AI applications. The study's reliance on English-language publications may have excluded valuable research from non-English speaking regions. Additionally, the bibliometric approach, while comprehensive for trend analysis, cannot capture the quality or practical impact of individual studies. The manual curation process, while improving accuracy, introduced potential subjective bias in relevance determinations.
Future Directions: The researchers recommend several critical areas for future investigation. First, there is urgent need for expanded K-12 and early childhood AI education research to address the current 89% higher education bias. Second, researchers should prioritize developing actionable ethical governance frameworks rather than merely identifying ethical concerns, addressing the current 78% framework deficit. Third, more longitudinal studies are needed to assess the lasting impacts of AI educational interventions beyond immediate outcomes. Fourth, increased inclusion of Global South perspectives and research contexts is essential for equitable AI education development. Fifth, stronger interdisciplinary collaboration between education and computer science fields should be fostered to bridge the current citation and collaboration gaps. Finally, researchers should focus on practical implementation studies that translate theoretical AI education concepts into effective classroom practices across diverse educational contexts.
Title and Authors: "Artificial Intelligence in Education (AIEd): Publication Patterns, Keywords, and Research Focuses" by Weijing Zhu (Guangxi Zhuang Autonomous Region Science and Technology Department), Luxi Wei and Yinghong Qin (Engineering Research Center for Intelligent Monitoring and Testing of Engineering Operation and Maintenance, College of Civil Engineering and Architecture, Guangxi Minzu University).
Published On: August 25, 2025
Published By: Information, published by MDPI