Automated Language-Based Feedback For Teamwork Behaviors

While most collaboration technologies are concerned with supporting task accomplishment, members of work teams do not always have the skills necessary for effective teamwork. In this research I propose that providing dynamic feedback generated by automated analysis of language behavior can help team members reflect on and subsequently improve their teamwork behaviors. This prospect is developed based on research in multiple disciplines, including teamwork effectiveness and social behaviors, feedback for training and regulating behaviors, and use of language in group conversations. To support this research, I directed the design and development of GroupMeter, a web-based chat system that analyzes conversations using a dictionarybased word count technique and visualizes indicators of language. I present a set of requirements for the GroupMeter system and the iterative process in which its design evolved. Findings from experiment 1 included a set of linguistic indicators that may serve as a useful source of automated feedback, such as agreement words and selfreferences, and that were embedded into the GroupMeter system. Experiments 2, 3 and 4 used GroupMeter as a research platform to examine the effects of automated linguistic feedback on team members. The experiments identify the conditions under which feedback positively enables reflection on and changes in language use and teamwork behaviors, as well as when it risks distraction and gaming behaviors. The findings are discussed in light of how feedback visualization shapes interpretations and perceptions of normative teamwork behaviors; ambiguity and benchmarking in representing social behaviors and language use; how to support balance between task-focus and socio-emotional interaction; and, improving teamwork behaviors versus “gaming the system”. This research contributes on three levels. Theoretically, it develops a threeway relationship between teamwork, feedback, and language, by tying together theories from multiple domains and supporting this relationship with empirical findings. Practically, it demonstrates a novel technique for training people to develop their teamwork skills. And design-wise, my work adds to the accumulating knowledge about groupware technologies that, while keeping the team activity in the center, illuminate peripheral awareness information about social interaction