Sit Down Now: Computationally Identifying Classroom Management Practices

Classroom management is essential to the work of teaching and learning, affecting teacher retention, school climate, and academic and behavioral outcomes for students. Yet, teachers face difficulties developing competencies in classroom and behavior management. Identifying and improving practices at scale is challenging as existing methods require expensive classroom observations by experts. We propose the task of using natural language processing to computationally identify management practices in classroom discourse. We release an annotated dataset of teacher utterances from observation transcripts and introduce two supervised machine learning approaches: binary classification of whether an utterance includes classroom and behavior management language, and prediction of the extent to which an utterance represents a punitive management attitude. Our automated language-based measures of classroom management show significant correlations with human-rated observational measures of instructional quality, student and teacher perceptions of classroom climate, student achievement, and demographic factors. We further leverage our measures to analyze lexical characteristics of classroom management language and variations in management over time. The performance of these methods indicates the potential for future automated evaluation and analysis of classroom management practices.