Structuring Peer Interactions for Effective Learning at Scale

Chinmay Kulkarni
Stanford, 2015
Massive open online classes (MOOCs) offer an opportunity to dramatically broaden access to education. However, dramatically broadened access also creates challenges. Classes enroll tens of thousands of students, all of whom participate remotely and asynchronously based on their own schedule. This large, asynchronous and remote access in turn makes it challenging to scale effective teaching techniques that rely on personal interactions between teacher and student, such as open-ended assessment and discussion, and rapid formative feedback. This dissertation brings the benefits of effective teaching techniques to massive online classes, by introducing computational systems that replace hard-to-scale teacher- student interactions with peer interactions. Because peer interactions rely on interac- tions between students, they can potentially scale to any classroom size. In this disser- tation first, I first study the causal mechanisms that lead to the learning benefits of classroom techniques like feedback and discussion. Then, I introduce interfaces that combine these operative mechanisms with the properties of online classes, such as mediated communication and the large number of students. This dissertation develops these ideas through two large-scale systems, PeerStudio and Talkabout, which target fast, revision-oriented feedback, and global-scale student dis- cussions, respectively. This dissertation also includes the first large-scale evaluation of a global peer-assessment system. PeerStudio uses the temporal overlap in student schedules at large scale so that stu- dents receive fast, revision oriented feedback from classmates at any time of day. Talkabout leverages the globally distributed student participation to create discussions where students speak with peers with diverse experience and viewpoints. Controlled experiments show both systems improve both studentsí learning experience and their grades. These systems, and the large-scale evaluations that led to their design, point to a future in which classrooms rely on the collective experiences of their students, and students around the world have access to education that is as effective as it is accessible.