Assessing Classroom Discourse with Epistemic Network Analysis

• Michael Robinson, Computer Science, Delaware State University

• Leila Barmaki, Computer & Information Sciences, University of Delaware

Back to back talk in the classroom is the core of how efficiently students can learn from their teachers and peers. Understanding the moment to moment teacher and student interaction could be helpful to improve classroom productivity. In this project, we applied Epistemic Network Analysis (ENA) to model the occurrence of the specific ways teachers and students talk during a discussion, like posing a question or justifying an answer. We chose the National Center for Teacher Effectiveness (NCTE) transcripts, containing 1660 different 4-5th grade hour long math lessons handpicked and recorded from approximately 50 schools and 300 classrooms. The purpose of this project is to develop and apply a structured coding scheme that captures the key discourse moves used by teachers and students, and to use ENA to reveal how those moves connect and pattern within lessons. . To do this, we have developed a hybrid codebook, combining theory-driven constructs from the NCTE frameworks consisting of codes (such as teacher uptake, pressing for reasoning, and student reasoning) with data-driven discourse functions observed in the transcripts (such as answer elicitation, explanation, feedback, and peer referencing).We manually coded 11 different talk moves such as “Teacher Elicit” and “Student Answer”.These talk moves were then used in the ENA webtool as well as an implementation of an ENA model in R. The model showed the high co-occurrence between the Teacher Elicit (Teacher asking questions) and Student Answer (students answers teacher questions) codes. This suggests that in a conventional classroom environment student answers to teacher questions are expected. In the future, we  will compare discourse networks across classrooms, for example, between higher- and lower-achieving classes to examine how patterns of mathematical talk relate to instructional context and student outcomes, potentially uncovering the co-occurance of talk moves and average math scores.