EAGER: Orchestrating Productive Collaboration Among Students in Mathematics with Multimodal Machine Learning

This project addresses the need for more effective teacher-oriented support tools for collaborative learning through the development of a machine-learning powered technical innovation called MathCollaborate. Almost every student in the United States is required to take Algebra 1, yet 40% of students did not achieve even the lowest proficiency level measured by the National Assessment of Educational Progress in 2019; and there is every indication that the pandemic made the situation even worse. Among students of underserved groups, the achievement levels are even lower. Providing more exciting and collaborative ways for students to engage in rich mathematics activities and discussions is a priority, but a challenge for many math teachers. The project activities supporting the development and research of MathCollaborate will help address these challenges by providing participating teachers with insights about students' math performance, engagement, and discourse, both online and in-person, enabling more focused math instruction. This project will help identify effective instruction and pedagogy around the use of collaborative activities in classroom settings, more broadly. This project seeks to support math teachers and students as they engage in collaborative mathematics activities and discussions by leveraging artificial intelligence and machine learning for online collaborative math learning. The intellectual merit of this project aligns to three overarching goals. First, this project will examine how teachers utilize and conduct collaborative work in their classrooms and explore how technology could be designed and implemented to better support teachers’ needs. Second, the project team will explore the types of collaborative paradigms that emerge as groups of students interact and discuss mathematics content. Within this, we will leverage multiple data sources to study the interactions and discourse exhibited by students during collaborative activities as well as how these correlate with learning outcomes. Finally, the project team will utilize what they learn in this project in conjunction with multimodal machine-learning methods to build detectors of productive and unproductive collaboration strategies. The team will develop these detectors into a functional prototype of MathCollaborate and examine its ability to support teachers’ orchestration of collaborative activities in their classrooms. 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.