Using Machine Learning for Cognitive Presence Detection in Asynchronous Online Learning

• In Event: Marriott Roundtable Session One
  In Roundtable Session: Instructional Technology SIG Roundtables (Table 8)

Abstract

This study aims to explore and develop a machine learning model to automatically detect individual students’ levels of cognitive presence they exert while participating in online discussions. To this end, we used the Practical Inquiry Model to analyze 1,360 discussion posts gathered from the multiple sections of a graduate-level online course. We have used the training data to develop a machine learning model. Especially, we tested a deep learning algorithm, Bidirectional Encoder Representations from Transformers (BERT) model, to find the most suitable hyperparameters. Our final model showed a 94% accuracy in detecting cognitive presence levels reflected on given discussion posts. Findings demonstrate the potential of AI to detect students’ cognitive presence and provide real-time feedback.

Authors

• Jinho Kim, Georgia State University
  Presenting Author

• Min Kyu Kim, Georgia State University
  Presenting Author