PI: Prof. Michael Bernstein, Melissa Valentine

Abstract

Design team collaborations often turn sour: the team starts out well, but descends into hostility and dysfunction. Little is known about the causes of design team fracture: was it inevitable given the people on the team, or was it the result of early comments and behaviors cascading into negative outcomes? We will (1) design an experiment to understand the causes of design team fracture, (2) develop a applied machine learning algorithm to predict fracture, and (3) introduce a tool to help avert fracture. Our experiment brings together design teams to pursue a complex design task online, scrambling usernames so that team members do not realize when they are working with the same people again. This experiment allows us to study whether fractures occur consistently with the same group of people. Second, we will use log data from this experiment to perform a natural language and network analysis to predict team fracture long before it occurs. Finally, we will develop an intervention to give team members early warnings when the team is at risk of fracture, testing whether this knowledge helps teams avert it. Together, this research seeks to understand the causes of design team failures and ameliorate its worst outcomes.