Notre Dame - IBM Research: Understanding students' emotions in online courses

This post was written by John Dillon, 2016 USAID | ND Global Development Fellow

In July, I co-presented at a Learning Analytics conference, Educational Data Mining (EDM), with my colleague Malolan Chetlur, from IBM Research, Bangalore. This presentation was part of a larger collaboration between Notre Dame (Alex Ambrose, PI) and IBM Research (Bikram Sengupta, PI). The conference was held in Raleigh, North Carolina. The weather was surprisingly cool, and Malolan and I both recommend a fabulous vegetarian restaurant called Fiction Kitchen.

One of the exciting aspects of this conference is that it brings together people from several different disciplines to study how students learn. Fields like Computer Science, Learning Science, Natural Language Processing, Psychology, Data Science, and Education, as well as research, teaching, and commercial perspectives are in attendance to talk about how to make learning better.

Our presentation focused on understanding student emotion in a Massive Open Online Course (MOOCs). In an online classroom you cannot see how students are feeling. You don’t know if one student is falling asleep or another is bored or confused. So the challenge facing online education ventures such as edX, Coursera, and Udacity is that online learning at scale lacks the personal and intimate feel of a traditional classroom. This is, perhaps, part of the reason why students tend to drop out of an online course at a much higher rate than in traditional learning settings. In our study, we asked students to self-report their emotion. We then use these self-reported emotions in conjunction with student behavioral clickstream data and Data Science and Machine Learning methods to better understand what emotions students experience in an online course context.

The payoffs of our research are answers to questions like these: Which emotions lead to fewer cases of student dropout? What types of content promote positive learning emotions? How do students’ emotions shift over the duration of the course? How do you use dynamic content to respond to how a student is feeling?

From Raleigh, I headed to Detroit and from Detroit on to Bangalore, India, where the IBM Smarter Education Team is based. Since July, we’ve been working on describing and predicting student emotion based on behavioral data. The eventual aim is a MOOC with cognitive capabilities, i.e. a MOOC which attends to, and responds to, a learner’s emotions in order to promote learning. This technology, in turn, will help solve global challenges of affordable and quality education at scale.

We would especially like to thank USAID, the Kaneb Center for Teaching and Learning, the John Kelly IBM Internship Program, the Notre Dame Office of Digital Learning, and IBM Research--Bangalore. Without the support of these bodies, this research would not have been possible.