Estimating Collaboration through Variation in Time Series from Members in Group Works
In successful group works, members engage in monitoring where they guess what other members are thinking. Since monitoring promotes self-regulated learning, many teachers try to introduce group works. This paper proposes a method to inform teachers on the fly of the monitoring status of students in group works using inexpensive and less invasive sensors. In the field of education, the method enables a teacher who supervises many student groups to direct their group works to successful ones. The method collects accelerations of multiple body parts and pulse waves from each member using sensors. The method uses the singular spectral transformation (SST) to detect the relative changes of each signal in the course of time. When a significant change takes place, it is considered the members get arousal. The method considers group members engage in monitoring when significant changes appear simultaneously among them. In an experiment to detect successful discussion to integrate ideas from the members into one feasible solution, we have obtained the accuracy of over 0.7. It indicates that the simultaneity of significant changes detected by SST is effective to estimate the monitoring state. In the experiment result, acceleration often worked better than pulse wave signals. It was also found that each member has a different role in the group. The behavior of each member varies with the role of the member. These results show that the method has potential to estimate the monitoring state. They also imply examination of the acceleration of each body parts would enable us to estimate the role of the members. The method allows one teacher to lead many groups to successful group works.