BIGDATA: IA: Automating Analysis and Feedback to Improve Mathematics Teachers' Classroom Discourse

University Of Colorado At Boulder, Boulder CO

Investigators

Abstract

This research project will develop and study an innovative application - TalkBack - for addressing a significant challenge in education: providing teachers with personalized feedback on classroom discussion strategies. The TalkBack application builds on advances in deep learning for natural language processing and speech recognition to automatically analyze classroom discussions and reliably generate information about specific classroom dialog between student and teacher that occur in active learning. This research will significantly extend existing machine learning from simple descriptions, and automated classification, of dialog to more complex descriptions, and automated classification, of dialog using deep learning. TalkBack will be a cloud application available to any teacher who has classroom video and who wants to improve active learning in the classroom. The TalkBack application will consist of three interrelated components: a cloud-based big data infrastructure for managing and processing classroom recordings, deep learning models that reliably detect the use of talk moves, and an innovative interface that provides teachers with personalized, feedback on their use of discussion strategies during individual teaching episodes and longitudinally over multiple episodes. Two user studies will be conducted to gather information from math teachers related to the design and impact of the application. These user studies will include a pilot study in year 2 (n = 20 teachers) and a field study in year 3 (n = 100 teachers). The TalkBack application will provide an exemplar for a new type of translational activity enabled by big data: the reification of existing, well-researched theoretical frameworks in deep learning models. Building on NSF's investment in research on talk moves, including the Accountable Talk and the IQA frameworks, this work demonstrates how analyses of teaching practices using these frameworks can be fully automated and scaled up to support large numbers of teachers longitudinally over time. Furthermore, this effort will demonstrate how a cloud-based infrastructure supporting the detailed analysis of classroom recordings using speech and language processing can be used to develop next generation learning environments (in this case, personalized feedback on teaching practices) and to uncover new insights into teaching practices at scale. Specifically, this research will provide unprecedented insight into the ways that classroom discussions and student participation changes as teachers develop and expand their use of talk moves over time. This study will develop the big data application, TalkBack, providing immediate and actionable feedback to teachers based on self-recordings of their mathematics lessons. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.