Machine Assessment of Student Discussion-Board Formal-Style Debates


  • Michael Glass Valparaiso University
  • Alexis Cooper North Carolina A&T State University
  • Jung Hee Kim North Carolina A&T State University



argumentation-based learning, computer-supported collaborative learning, computer-supported discussion board, class debate


This paper explores machine identification of argumentative moves in asynchronous online student debates. Student debate is a technique for engaging with topics which have no clear answer. Online debates differ from less-structured discussion-board class discussions in that students are restricted to formal debating moves such as advancing a new argument, providing evidence, or rebutting an argument. They are also assigned to debate a particular side. This research explores whether the different argumentative skills in student debate can be distinguished from each other by reading the texts of the students’ discussion. The data for this research are the postings from 20 week-long student debates, an assignment in an educational technology class. Utilizing exclusively the text the students wrote, machine classifiers were trained to recognize the argumentative role of a message. If machines can detect an average difference between these debate moves from the text of the debate, and if they can detect the side of the debate, then the students are likely exercising different argumentative skills. The experiment shows it is possible to identify the skills which make debating different than free-form discussion with accuracy significantly better than chance.

Author Biographies

Michael Glass, Valparaiso University

Assoc. Prof., Dept of Computing and Information Sciences

Jung Hee Kim, North Carolina A&T State University

Assoc. Professor, Computer Science Dept., (ret.)