Predicting Roles and Knowledge Construction Processes in Asynchronous Discussion Using Machine Learning

• In Event: AERA Roundtable Session Saturday 3:20 pm Four Seasons Ballroom 1
  In Roundtable Session: AI-Driven Innovations in Instructional Technology (Table 5)

Abstract

This study investigates the use of Machine Learning (ML) to predict participant roles and knowledge construction processes in asynchronous collaborative learning. Utilizing data from an online Information Systems course, we applied TF-IDF and Part of Speech tagging to transform text into feature vectors. Logistic regression models predicted roles such as Summarizer, Skeptic, and Source Searcher with high accuracy. However, predicting the Moderator and No Role categories was less accurate. The models also effectively identified knowledge construction processes like Conflict Consensus and Externalization. The findings suggest ML's potential to support instructors in managing collaborative activities and enhancing learning experiences. Further refinement of these models is needed for broader educational applications.

Authors

• Minkyung Lee, Pennsylvania State University

• Priya Sharma, Pennsylvania State University