Abstract
This study aimed to 1) identify latent research topics in educational measurement and evaluation from academic databases covering the years 2008–2025 using Structural Topic Modeling (STM), 2) analyze topic prevalence trends and forecast their growth trajectories for 2027 using Bayesian Multilevel Piecewise Regression (BMPR), and 3) synthesize strategic research themes from the topic classification and forecasting results. The dataset consisted of 1,695 research articles from the proceedings of three international conferences. The data were processed to extract 10,858 standardized keywords and key phrases, which were then analyzed using STM to identify latent research topics. Topic labeling and interpretation were conducted through a human-in-the-loop process in collaboration with large language models. The findings showed that STM identified 95 latent research topics, which could be semantically organized into two major domains: assessment, evaluation, and policy; and innovation and technology for learning. BMPR results revealed high variability in topic prevalence trends, reflecting significantly different patterns of growth, slowdown, or stability across topics. Forecasting and topic screening identified 40 high-potential topics, which were subsequently grouped into 13 subclusters and synthesized into four strategic research themes: AI-driven learning systems, measurement and validity in large-scale assessment, equity and engagement, and responsible use of artificial intelligence.
Publisher
Faculty of Education, Chulalongkorn University
DOI
10.14456/jescu.2026.5
First Page
1
Last Page
24
Recommended Citation
Srisuttiyakorn, Siwachoat; Sriklaub, Kanit; and Saifah, Yotsawee
(2026)
"Forecasting Trends in Educational Measurement and Assessment Topics in the AI Era : Text Mining Analysis with Bayesian Multilevel Piecewise Regression Model,"
Journal of Education Studies: Vol. 54:
Iss.
2, Article 4.
DOI: 10.14456/jescu.2026.5
Available at:
https://digital.car.chula.ac.th/educujournal/vol54/iss2/4
