As the workplace has become increasingly diverse, there has been a tension between the promise and the reality of diversity in team process and performance. The optimistic view holds that diversity will lead to an increase in the variety of perspectives and approaches brought to a problem and to opportunities for knowledge sharing, and hence lead to greater creativity and quality of team performance. However, the preponderance of the evidence favors a more pessimistic view: that diversity creates social divisions, which in turn create negative performance outcomes for the group. Why is the reality of diversity less than the promise? Answering this requires understanding a variety of factors, including how diversity is defined and categorized, and the moderating as well as mediating processes that affect the diversity–process–performance linkage. We start with a definition. The word diversity has been used to refer to so many types of differences among people that the most commonly used definition—“any attribute that another person may use to detect individual differences” ( Williams & O'Reilly, 1998 , p. 81)—while accurate, is also quite broad. As a result, various categorization schemes based on factors such as race or gender, or based on proportions such as the size of the minority, have been used to further refine the definition of diversity in teams. The choices researchers have made in using these categorization schemes, however, do lead to particular trade-offs. Factor approaches, for example, allow an examination of multiple types of diversity and the interactions among them but ignore the sizes of factions and subgroups. Proportional approaches allow the consideration of minority-group size, and hence the study of issues such as tokenism, but also tend to focus on only one type of diversity and thereby overestimate its relevance relative to other types. The underlying effects of diversity, whichever way it is defined and categorized, have typically been understood through three primary theoretical perspectives: the similarity–attraction paradigm, self- and social categorization, and information processing. These approaches also have their biases. The predictions of similarity–attraction theory are straightforward: Similarity on attributes such as attitudes, values, and beliefs will facilitate interpersonal attraction and liking. Empirical research has supported that surface-level similarity tends to predict affiliation and attraction. The similarity–attraction paradigm was developed to understand dyadic relationships. Yet, individuals can express preferences for membership in particular groups even when they have had no prior social interaction with members of that group. This is primarily a cognitive process of categorization: Individuals are postulated to have a hierarchical structure of self-categorizations at the personal, group, and superordinate levels. Research has demonstrated that the specific categories on which we tend to focus in categorizing others—such as race, gender, values, or beliefs—are likely to be those that are the most distinctive or salient within the particular social context. The act of social categorization activates differential expectations for in-group and out-group members. This distinction creates the atmosphere for stereotyping, in which out-group members are judged more stereotypically than in-group members are. The self-categorization/social-identity and similarity–attraction approaches both tend to lead to the pessimistic view of diversity in teams. In these paradigms, individuals will be more attracted to similar others and will experience more cohesion and social integration in homogeneous groups. The information-processing approach, by contrast, offers a more optimistic view: that diversity creates an atmosphere for enhancing group performance. The information-processing approach argues that individuals in diverse groups have access to other individuals with different backgrounds, networks, information, and skills. This added information should improve the group outcome even though it might create coordination problems for the group. As we disentangle what researchers have learned from the last 50 years, we can conclude that surface-level social-category differences, such as those of race/ethnicity, gender, or age, tend to be more likely to have negative effects on the ability of groups to function effectively. By contrast, underlying differences, such as differences in functional background, education, or personality, are more often positively related to performance—for example by facilitating creativity or group problem solving—but only when the group process is carefully controlled. The majority of these effects have typically been explained in terms of potential mediators such as social integration, communication, and conflict. However, the actual evidence for the input–process–output linkage is not as strong as one might like. Clarifying the mixed effects of diversity in work groups will only be possible by carefully considering moderators such as context, by broadening our view to include new types of diversity such as emotions and networks, and by focusing more carefully on mediating mechanisms. As we delve into advice for organizational teams to enhance the assets of diversity and manage the liabilities, we focus on the benefits of “exploring” as opposed to “exploiting” types of tasks, of bridging diversity through values and goals, and of enhancing the power of the minority. Finally, we end with suggestions for how organizations can learn to create incentives for change within the firm.
