Examining the explanation: AI-based classification of explanatory textual sequences in secondary mathematics teaching

Abstract The role of explanation in mathematics teaching is central to promoting student understanding, yet research has been limited by the challenges of analysing large volumes of classroom discourse. This study explores the use of AI to identify, segment, and classify explanatory textual sequences in secondary education mathematics classrooms. Drawing on textual linguistics and typologies of explanation in mathematics education, the study analyses 28 h of classroom recordings from six high school mathematics teachers. A two-step methodology was applied: (1) segmentation of explanatory sequences using Whisper and GPT-4, and (2) classification into interpretative, descriptive, and justificative types. The inter-rater agreement between GPT-4 and the coding team reached 92.4% for segmentation and 77.5% for classification, exceeding the accepted reliability thresholds. Correlation analyses suggest a negative relationship between interpretative and justificative explanations. The findings support the potential of AI for large-scale discourse analysis in educational settings and highlight the value of combining linguistic theory with AI to better understand mathematics teaching practices. Implications for teacher training, discourse-based instruction, and future AI integration in mathematics education research are discussed.