Prediction of Design Team Performance through Analysis of Language Use
The subject of how teams work together to negotiate and accomplish things has long been studied in several fields, including psychology, engineering, management, and anthropology. In the last few decades, researchers have increasingly focused on ways team members synchronize and mirror each others' speech and movement in meetings and conversations. People acting this way towards each other are typically not aware of, or little aware of, this aspect of their behavior. Related phenomenae of interest go by such terms as interpersonal adaptation, nonconscious mimicry, behavioral matching, the chameleon effect, and linguistic entrainment. These have been shown in several ways to affect how people interact; to be correlated to such things as trust, rapport, engagement, and liking; and to be a robust way to examine interaction.
A simple way to assess lexical linguistic entrainment has been developed called the Language Style Matching (LSM) measure (Gonzales, Hancock, & Pennebaker, 2010). In the research to be presented, some simple statistical signal processing techniques are used in conjunction with LSM to develop several new derived measures that are predictive of design team performance. One result is the discovery of an anti-entrainment phenomenon that is associated with better-rated design work in short, brainstorming-style design meetings.
Once a transcript of a meeting is available, these measures are largely automatable, and are topic-independent. This promises to make it much more cost-effective to examine a large number of design- and task-oriented conversations and robustly compare them, as a useful window into group dynamics. The presentation will further discuss how these results are consistent with and expand on previous research, and what limitations apply to this way of looking at teamwork.