The Effect of Role Specialization And Transactive Memory Systems On Performance in Data Science Teams

Thumbnail Image


Publication or External Link





Teamwork is an integral part of data science work. Data science work requires knowledge from many different disciplines including statistics, information visualization, programming, and subject matter knowledge related to a given set of data sets (e.g., politics, education). Data science teams are often formed by individuals who have different areas of knowledge and expertise and, as a result, may take on different functional roles within a team. Due to their distinctive expertise, members in data science teams may take on specialized task roles matching their expertise, and such division of labor could increase coordination cost among team members. As data science work is often open-ended and dynamic by nature, high coordination costs could deteriorate performance in data science teams. In this research, I argued that developing shared cognition on who-knows-what (i.e., transactive memory system, abbreviated as TMS) in data science teams would be beneficial for team performance, especially when the members have specialized roles. I conducted two studies to understand the effect of role specialization and transactive memory systems on team performance with a goal to identify and test a lever to facilitate transactive memory system in data science teams. I collected data from two consecutive Data Challenge events; Data Challenge is an week-long data science competition hosted annually as a university-wide event. In Study 1, I conducted an observational study by collecting survey data from 74 individuals in 36 teams in Data Challenge 2019. In Study 2, I conducted a field experiment to examine the effectiveness of an experimental intervention designed to facilitate transactive memory system in data science teams by highlighting any inaccuracies in the perceived expertise between members.