Towards the Automatic Assessment of Student Teamwork
Association for Computing Machinery
Teamwork skills are crucial for college students, both at university and afterwards. At many universities, teams are increasingly using discussion platforms such as GroupMe and Slack to work virtually. However, little has been done so far to understand how to use the data these platforms generate to analyze student teamwork behaviors, and so to support or improve those behaviors. Furthermore, these data have not been exploited to determine whether effective student team members share any other traits. This project therefore attempts to determine (a) whether there are any characteristics common to the online discussion behaviors displayed by high-performing vs non high-performing student team members and (b) whether high-performing vs non high-performing student team members share any apparently teamwork-exogenous attributes. We find that the features of team member communication that best predict team member performance are sentence length and the number of words contributed to the team's discussion, with a range of other features playing a smaller role. We also find that teamwork-exogenous factors (such as pre-college ACT score, and number of credits attempted during the semester) were only moderately predictive of team member performance.