What teachers discuss and how they perceive and feel: Topic modeling and sentiment analysis of a generative artificial intelligence learning group on facebook

Rapid advancements in generative artificial intelligence (GenAI) have been hailed for their potential to transform teaching practices, student achievement, and teachers’ administrative work. However, previous studies have primarily relied on qualitative interviews, surveys, or experimental validations of specific tools’ effects on achievement. Therefore, the full range of teachers’ concerns about learning and applying GenAI has yet to be widely explored using a quantitative, big-data approach. To bridge that gap, this study employed latent Dirichlet allocation for topic modeling, in combination with sentiment analysis, on 96,693 posts from the largest GenAI teacher-learning group on Facebook. This resulted in four (K = 4 topics): 1) using GenAI for teaching and administrative work, 2) using GenAI for student development, 3) tips for GenAI usage and assessment, and 4) GenAI ethics and policies. Teachers in this community showed relatively positive sentiment toward the first and third topics, but more neutral sentiment toward the second and fourth ones. Methodologically, this study’s integration of topic modeling with sentiment analysis to analyze large-scale teacher discussions provides a bottom-up perspective that exceeds the possibilities of controlled experimental settings. The findings derived from that approach, in combination with qualitative analysis of posts, add theoretical depth to scholarly understanding of the emotional and role-based factors influencing GenAI integration by linking teachers’ positive sentiments explicitly to perceived professional transformations. And practically, the findings offer insights for tailoring GenAI-related training, supporting peer mentoring, refining assessment of students engaged in human-AI collaboration, and formulating ethical guidelines that align with teachers’ real-world concerns.