Creating Links: Building an Educational Platform to Ask Relevant Questions in Education

Abstract

In this thesis, I document the findings and process through which we built an educational platform (JANN) to do research while having a positive impact on a community. Through JANN we have coordinated more than 100k hours of tutoring sessions and built (to our knowledge) one of the largest databases of educational recordings in the world. Broadly the contributions here are twofold: first, we demonstrate the research potential building a platform can offer. Second, using our educational platform, we pursue novel questions in the field of education with granular information that is traditionally inaccessible for research.

After introducing the work and describing the construction of the platform, the first chapter details an RCT where we show the effect of receiving tutoring on Math performance. Second, we document how we built an estimator of emotions using audio. The estimator was further validated on our dataset and then used to show that activating emotions are related to better class quality. Third, we document an RCT where Math tutors were asked to dedicate some time per week to teach Socioemotional learning skills. We show that this had a positive effect on learning. Moreover, it also caused tutors to teach longer Math classes. Students showed more trust in their tutors, and ultimately the classes had a higher prevalence of positive emotions. Finally, we also study doing causal inference on observational data on another platform. Using Facebook data we study digital groups and through a regression discontinuity design we find that joining a group has a positive effect on making new friends and can diversify a person's connections in terms of income.

Overall, we find that building a platform, can broaden the granularity of the data one has access to, make research more scalable, and ultimately also have a positive effect on a community.