Simulation of the Categorization-Elaboration Model of Diversity and Work-Group Performance

The relationship between the diversity of work-groups and their performance continues to be a key concern in the study of organizational behavior. Several models have been proposed to explain this relationship, generally concentrating on the interplay between two main factors: diversity as a source of varied knowledge and viewpoints that a group can draw upon to increase its performance, and diversity as a source of dissention in groups, causing group fracturing and bias, leading to decreases in performance. Recently a model called the categorizationelaboration model (CEM) (van Knippenburg et. al. 2004) was proposed which integrates existing research in diversity and group performance into a unified framework. We perform an agent-based simulation of the CEM where groups are modeled as coalitions of rational agents which draw from distinct experience pools and which collectively try and solve a simple forecasting problem. We simulate how the performance of the coalition varies with the diversity of the agents' background experiences, and find that the resulting performance/diversity relationship is curvilinear in nature (specifically, inversely u-shaped), as predicted anecdotally in the van Knippenburg work. Additionally, we find a point of unstable equilibrium in the performance/diversity curve at the no-diversity point, such that at the no-diversity point, small increases in diversity have little or no effect on performance. We point out a connection between the existence of this feature, which would seem to highlight the importance of external diversity-encouraging efforts such as affirmative action-type initiatives and early economic work which suggests that market-based forces should be sufficient to ensure high levels of diversity in organizations.