Forecasting Student Achievement: The Potential of Natural Language Processing

• In Event: Poster Session 3
  In Poster Session: Learning Process and Effects of Technology-Enhanced Learning

Abstract

Educators have long desired to be able to predict student achievement before it is too late to do anything about it. The existing literature on predicting achievement focuses mostly on demographics and typically does not control for students' intentions. The current research presents a prediction model to anticipate course completion among students in three massive open online courses (MOOCs), who intend to complete their course. Using a machine learning Natural Language Processing (NLP) approach to analyze students’ unstructured text responses to a utility value prompt in pre-course surveys, we find that a NLP model can add predictive power over and above other standard approaches. The results highlight the potential of using unstructured language data to predict student success.

Authors

• Carly Dolan Robinson, Harvard University

• Mike Yeomans

• Justin Fire Reich, Massachusetts Institute of Technology

• Chris S. Hulleman, University of Virginia

• Hunter Gehlbach, University of California - Santa Barbara