Using decision trees to disentangle the complex relationship between diversity and knowledge sharing

Abstract

Much of the literature in the field of diversity serves to disentangle the seemingly conflicting impact of diversity on performance. While nonlinear models and models of external factors that either mediate or moderate the relationship have significantly advanced the field, we examine the use of decision trees as a novel way to model the relationship between diversity and knowledge sharing -- a recognized antecedent of performance. In order to demonstrate the power of decision trees, we study five types of visible diversity (age, gender, education level, religion, and region of origin) along with one measure of personality (anxiety). Using data collected within an academic institution, we show that decision trees can automatically find demographic faultlines that serve to predict levels of both explicit and implicit knowledge sharing.