Large group collaboration

This project studied large group collaboration by identifying "action contributions" of each individual in a group from their body postures. The project used OpenPose algorithm and clustering to decipher actions from video data, which is especially useful in the absence of think-aloud protocols and post-task interviews. The OpenPose algorithm detected skeletal postures of large groups (12-14 visitors/group) of co-located museum visitors collaborating on a problem solving task. The groups were tasked to attain diversity in Connected Worlds and sustain it for a period of time. We further used clustering to assign action labels to the poses and define contribution "actions" in the collaborative task and understand division of labor within the group.