SBIR Phase I: Positive Effects of Feedback and Intervention for Engagement in Online Learning

This SBIR Phase 1 project will study how communications feedback loops effect online learning, especially in group settings. The explosion of online learning experiences in recent years, in which many courses fail to engage learners or help them achieve their academic and professional goals, has put focus on the problem of engagement. Interest in online learning remains high, as it presents an opportunity to broaden the reach of institutions and companies, but success rates in broadly available learning experiences are low. Built upon foundational research in computational social science, the project will use specialized video and text chat tools to collect data from groups engaged in synchronous online learning, analyze and categorize patterns of communication, and give feedback to students to help them have better, more successful experiences online. In keeping with NSF's mission to promote the progress of science and advance national welfare, the project applies emerging innovations in artificial intelligence to the problem of delivering high-quality learning at scale. While most applications focus on individual learners and creating new personalized educational pathways, this project highlights the importance of collective learning and the power of peer engagement to help all members of a learning community reach their goals together. This project introduces several innovations in data capture and analytics, as well as feedback to learners while they are collaborating online. While many students use online collaboration tools such as discussion boards, chat, and video, this project focuses on providing interventions to change behavior in real-time. During conversations, such as those happening when people brainstorm to solve a problem or work on shared assignments, certain vocal patterns indicate engagement, reveal unstated agreements and discords, and expose individual biases. Using deep learning and data modeling, real-time feedback about conversational dominance is generated and shown to learners while they are speaking. The project further explores the effect of machine-generated insights about what happened during the collaboration, and makes recommendations about behavior modifications, based on predictive models. For more nuanced insights, these data are paired with facial-gestural data, such as nodding or raising one?s eyebrows. Together, these innovations combine data that has never been collected at scale before with social science data models uniquely applied to online learning. The goal of the research project is to prove that insights like these have a net positive effect on learning outcomes and raise the level of satisfaction with online learning experiences that incorporate these tools. 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.