Evaluating the Feasibility of Bayesian Knowledge Tracing for Online Instruction

• In Event: Online Teaching and Learning SIG Paper Session 1

Abstract

We analyze two potential applications of a statistical inference methodology, based on Bayesian Knowledge Tracing (BKT) and discuss the value of this approach for online learning platforms. One application applies this methodology to a comparison of forms of guidance in an online collaborative inquiry-based science unit for middle schoolers. The other application investigates recommending teacher resources based on probabilistic learning gains from different videos and problem sets. We identify potential questions that could be answered through these types of applications and investigate the feasibility of applying these methodologies to new domains. We discuss how applying these findings can lead to personalization of resources to create more efficient and effective learning experiences.

Authors

• Hannah Gogel, University of California - Berkeley

• Elizabeth McBride, University of California - Berkeley

• Jonathan Michael Vitale, University of California - Berkeley

• Zachary Pardos