IDENTITY DIVERSITY
A man is like a bit of Labrador spar, which has no lustre as you turn it in your hand, until you come to a particular angle; then it shows deep and beautiful colors. There is no adaptation or universal applicability in men, but each has his special talent, and the mastery of successful men consists in adroitly keeping themselves where and when that turn shall be oftenest to be practised.
—RALPH WALDO EMERSON, Experience
Diversity bonuses result from differences in what we know, how we perceive the world, the frameworks and models we use to organize our thoughts, and the ways we generate ideas. The notion of a cognitive repertoire is an artificial construct. How we think about a problem, make a prediction, or evaluate a strategy depends on complex interactions in our brains within a network of neurons, axons, and dendrites.
A variety of factors cause people to build diverse cognitive repertoires. Nature must play a role, but for many of the contexts covered here, our experiences and formal training matter more. A doctor possesses a cognitive repertoire distinct from that of a plumber or a materials scientist more because of training and experience than genetics.
In this chapter, I take up the question of the extent to which identity differences contribute to relevant cognitive diversity. To the extent that they do, identity differences contribute to cognitive diversity bonuses.
Identity diversity commonly refers to differences in race, gender, age, ethnicity, religion, physical qualities, and sexual orientation. Logically, a bright line separates this from cognitive diversity. One corresponds to how we think, while the other corresponds to categories we use to define people. Empirically, the distinction becomes blurry. Our identities influence what we know, how we perceive events, and how we think.
To connect identity diversity to cognitive diversity requires unpacking the term identity. I present three frameworks that can organize our thinking: the icebergs, the timber-framed house, and the cloud. The icebergs (there will be two) highlight the fact that we see some attributes while others lie below the surface. The timber-framed house represents identities as consisting of connected components that combine to form a whole. We cannot decompose a Japanese American woman’s identity into a Japanese part, an American part, and a female part. The cloud emphasizes the variation within identity categories.
All three frameworks prove central to the main argument of this chapter. When we try to link identity diversity to cognitive diversity, we lean on categories. Those categories are based on what we see above the iceberg’s waterline. They structure the questions we ask and the inferences we draw. Claims that women think differently from men or that African Americans bring a unique perspective rest on crude categorizations. Those categories slice off single dimensions from the timber-framed house. Pulling off just one attribute and drawing inferences will produce errors. And last, no matter how finely we make the categories, within-category variation will remain.
As a starting point, we must keep in mind that any group, even a group whose members possess the same identity attributes, will be cognitively diverse because no two people possess identical cognitive repertoires: no two people possess the same information, apply the same perspectives, or carry around identical collections of mental models. That is true even of people who belong to the same identity group.
However, given that our identities influence how we construct our lives and how others treat us, we would expect identity-diverse groups to be more cognitively diverse than homogenous groups. We would also expect people to behave differently when in a diverse group. Empirical evidence supports both inferences.1
Given the importance of identities, they must correlate with how we think. In some domains, the links are obvious. Adding race and gender diversity would enhance a group discussion of the Roe v. Wade decision, the significance of Harper Lee’s To Kill a Mockingbird, or the Barack Obama presidency. On the other hand, identity cannot have superordinate influence in all cases. Teams of materials scientists developing organic solar cells lean more on educational and experiential diversity than on identity diversity.
Though our identities may play a smaller role in driving scientific understanding and discovery than in interpreting the law or making psychological inferences, they may still have an impact in science through the analogies we invoke. Think back to high school biology when you had a test on the various parts of a cell and their functions. One of the cell’s parts, the Golgi apparatus, packages proteins within the cell’s vesicles and sends them on their way. To recall and explain the Golgi apparatus’s function, students invoke any number of analogies—a post office, a sugarcoating factory, a baggage department, or a finishing school.2 We can connect these analogies to life experiences. Few rural kids would think up the finishing school analogy.
The choice of analogy frames how someone thinks of the Golgi apparatus. The student who thinks of the Golgi apparatus as assigning tags on luggage understates the organelle’s function. In her view, the organelle puts proteins in boxes and slaps on addresses. Conceptualizing it as a finishing school hints at a more elaborate function and could spur deeper inquiry.
The investigation of the connections between identity and cognition will not produce definitive answers. Evidence that identity diversity correlates with or causes cognitive diversity need not imply that identity-diverse groups always make better choices, come up with more innovative solutions to problems, make more accurate predictions, or elaborate more creative alternatives. In some cases, identity-diverse groups might perform best. In other cases, identity diversity might add little relevant cognitive diversity.
Identity will prove a complex and fluid combination of attributes that cannot be captured by a set of boxes on an application form or variables in a statistical regression. That complexity undermines any simple causal explanation. We cannot prove that identity diversity creates beneficial cognitive diversity. What we can do is explore the possible linkages and gauge their plausibility.
AN IDENTITY PRIMER
When describing identity differences, laypeople and scientists alike rely on six primary categories: race, gender, age, ethnicity, physical capabilities, and sexual orientation. More expansive categorizations include neighborhoods, family structure, diet, forms of artistic expression, social norms, education, socioeconomic status, genes, ancestry, dialect, and power relations. We might describe a friend Gunther as a black, German, heterosexual, Catholic male. These categories differ in their permanence. Gunther will always be black and of German descent. His religion or sexual orientation may change.
Identity categories reduce people to a handful of tags—white, male, differently abled, and so on—that cannot capture in full the rich diversity of individuals they contain. We would categorize some of the world’s most gifted artists and athletes as disabled. Beethoven became deaf. Jim Abbott, born without a complete right hand, quarterbacked his high school football team, won the Sullivan Award as the nation’s top amateur athlete, and pitched a no-hitter for the New York Yankees.
Despite obvious flaws, identity categories persist. They enable meaning making. They function as a self-reinforcing accounting standard. With them, we can describe human diversity, identify discrimination, and measure inequality across groups.
In the past, people thought of identity attributes as essential.3 Essentialism assumes that identities correspond to innate, unchanging characteristics. Men were believed to possess an innate male essence observable through characteristics. Male characteristics, both physical and cognitive, were thought to differ from the innate characteristics of women, who were thought more caring and sensitive. Similarly, the innate characteristics of Asians were thought to differ from those of Europeans and Africans. Asians were seen as more collectively oriented.
Essentialism derives from Aristotelian foundations: the essence of an orange differs from the essence of a peach, and the essence of a dog is distinct from the essence of a badger. Essentialism predates genetics. Many categories once deemed essential have been found to lack genetic foundations. Racial classifications are a notable example. People who identify as African American can trace, on average, one-fourth of their genes to Europe.4
Given that people’s beliefs about their racial identity and the identities projected on them by others matter more than genetics, we now think of race as socially constructed.5 Society creates the categories and endows them with meaning.
An analysis of any particular trait reveals the challenges of essential designations. Height satisfies two properties of an essential attribute: it is largely determined by genetics and it varies across groups. Data based on a quarter million people reveals seven hundred genes spread over four hundred gene regions that can explain 80 percent of height variation.6
We also see variation across groups. The average height of Dutch men (over six feet) exceeds that of American men by about three inches. That statistic will not surprise anyone who has wandered Amsterdam’s Schiphol Airport.
And yet, the tallness among the Dutch is at best temporally essential. Over the past two hundred years, the average height of a Dutch male has increased more than eight inches. Two centuries ago, the Dutch were shorter than Americans. If being tall is now an essential trait of the Dutch, then, in the recent past, being short was an essential trait.7
Of course, height is just one characteristic of the Dutch. The Dutch are also thought to be less generous (“going Dutch”) and more open-minded. Those attributes seem less likely to be essential than height. If we stereotypically categorize the Dutch as a tall, frugal, open-minded people, we surely err.
By applying frameworks, even at the level of analogy, we can think about identity with greater subtlety and avoid these stereotypes. We are also better able to think about how and when identity diversity might connect to cognitive diversity.
THE ICEBERGS
The first framework, the icebergs, distinguishes attributes that compose identity diversity—race, gender, ethnicity, sexual orientation, age, and physical capabilities—by their observability (see figure 4.1). We see a person’s skin color, gender, and age. We hear her dialect. We see physical qualities. We do not see values, beliefs, religion, history, or ancestry. The traditional identity iceberg analogy categorizes attributes as either above or below the waterline.8 We observe those attributes above the waterline. We can only infer attributes below the waterline.
The iceberg analogy reveals an inherent selection bias in impressions of identity diversity. We construct categories and make inferences based on (or biased toward) observable identity characteristics. Many lump Korean Americans, Chinese Americans, and Japanese Americans into a single category: Asian Americans. That categorization sweeps ancestry, ethnic identity, and religious practices under the rug. The media emphasizes the disproportionate number of Asian Americans at elite colleges and universities. The large number of evangelical Asian American students goes unnoticed.9
Figure 4.1 An Iceberg Representation of Identity Diversity
If selecting a representative group of students to discuss their experiences at Berkeley based on observable racial characteristics, we might choose two Asian Americans. If both students identify as Chinese American, the two of them may not be that diverse. A more deliberate set of choices, one that looks below the waterline, might select a first-generation Chinese American Buddhist and a third-generation Korean American Christian. This second pair would produce a broader, more nuanced discussion.
The iceberg framework also highlights a distinction among perceived, expressed, and internal identifications. A person may internally feel as though they belong to one identity group, express a different identity through behavior, and be perceived as belonging to either of the two identity groups by different people. For example, self-reported racial categorizations among multiracial adolescents depend on context. Their answers to questions about their race at home in the presence of their parents differ from the answers they give at school when surrounded by friends.10
This traditional iceberg model captures impressions of identity in the physical world. The virtual world creates a second iceberg. Data gathered from the web reveal attributes that we may not observe at street level. The virtual iceberg includes criminal records, property ownership, and political donations, none of which we can infer by passing someone on the street.
The differences between these two icebergs have become more relevant because social scientists and human resource professionals increasingly evaluate the effects of diversity by analyzing online data; in other words, the virtual iceberg. A social scientist working with census data might include a person’s neighborhood and social class, two attributes that lie below the physical waterline. Social scientists then measure relationships between neighborhood characteristics and performance in school, probability of criminal behavior, and likelihood of drug use.
As more attributes rise above the virtual waterline, we can better avoid the stereotypes we jump to in the physical world: African Americans are more likely to go to jail, Asian Americans perform better in school, white Americans commit more acts of terrorism. Those correlations may be statistically valid, but they lack causal explanations. Any explanatory attributes for criminal behavior or educational performance, such as income, social class, parental education, neighborhood, religious affiliation, or mental health, likely lie below the physical waterline.
To categorize someone as Asian American and make inferences based on that categorization can lead to wrong actions as well as improper inferences. I once met a third-generation Japanese American who had been hired to manage a company’s Asian American client list. The company hoped his shared identity would allow him to build trust with clients, improve communication, and increase revenue.
Upon being hired, he found that the company’s Asian American clients were Korean Americans. He soon quit, not because he was not successful at his job; he performed above expectations. He left because he believed that any company with such a poor understanding of its clients would not be successful in the long run.
THE TIMBER-FRAMED HOUSE
The second analogy, the timber-framed house, represents a person as possessing multiple, connected identity attributes.11 Just as two parts of a timber-framed house can be nailed together or lie far apart, so too can pairs of identity attributes. A person’s neighborhood and social class connect more closely than her skin color and diet, while her ancestry probably links more tightly to her religion than to her current social status.
The timber-framed house analogy warns against separating out effects of individual identity attributes. A person’s entire identity influences her life experiences and therefore her cognitive repertoire. We cannot and should not think that a person’s identity-based influences can be decomposed into a sum of effects from individual identity attributes.12 The identity influences on the cognitive repertoire of an African American woman cannot be decomposed into an African American effect and a woman effect.
Figure 4.2 Identity as a Timber-Framed House (Sen and Wasow, “Race as a ‘Bundle of Sticks’ ”)
Therefore, we cannot gather data on the opinions of African American women on a proposed educational reform and tease out a woman effect and an African American effect. The effects of gender and race interact and intertwine. A person’s identity consists of the whole structure. We can no more understand a person by listing her identities than we can appreciate a house by taking an inventory of the materials used in its construction. We need to know how the parts interact.
This nonadditivity of attributes arises in intersectionality theory.13 Intersectionality teaches us that considering the effects of race and gender separately can obscure discrimination and miss forms of oppression. For example, a company that hires white and black men and only white women might have representative numbers of blacks and women, even though it hires no black women.14 Alternatively, a company that hires only white men and black women could have more than a representative number of blacks yet no black men. Relatedly, the discrimination experienced by black women may differ markedly from that experienced by women or blacks.
The inseparability of attributes implied by the timber-framed house analogy should give us pause when interpreting data sorted by single attributes. Single-attribute correlations obscure richer stories. For instance, women constitute approximately three-fourths of veterinary students. We cannot infer that being a woman makes one more likely to want to be a veterinarian. The vast majority of women veterinary students are white. Therefore, it is not women who pursue veterinary degrees, it is white women.
The timber-framed house framework is important to keep top of mind because we are prone to crude, inaccurate inferences based on single attributes. These are often wrong. Yes, women make up two-thirds of fund-raisers, but few of those women are Latina. Yes, two-thirds of judges are men, but not many are Asian American.
THE CLOUD
When we invoke a crude category—European American, African American, and so on—we condense diverse populations to a single point. This diversity of cognitive repertoires within any category reveals the impossibility of asking any one person to represent an identity category.15 This observation leads to the third framework: the cloud. Any short list of identity attributes will be shared by a diverse set of people. The identity bundle that includes the features educated, heterosexual, Catholic, and Irish American is held by a diverse collection of people including comedians Conan O’Brien and Bill Murray and television correspondent Elizabeth Vargas.
Within any bundle, people differ in diet, skin color, social class, political views, and family structure. That holds true for any identity bundle. Categories do not divide people into clean sets of identical people; they create neighboring clouds.
The variation with the cloud also problematizes inferences. We cannot claim that someone brings a woman’s perspective. There is no such thing. The set of women is too large and diverse to have a unique shared perspective. The most we could say is that the set of ways in which women look at a particular problem differs from the set of ways in which men frame it.
As we move forward, we must keep all three analogies in mind: identity consists of multiple dimensions, some of which we see and others of which we don’t (the iceberg); these attributes cannot be separated because they connect (the timber-framed house); and within any category, we find a diverse group of people (the cloud).
CONNECTING IDENTITY DIVERSITY TO COGNITIVE DIVERSITY
We are now in a better position to explore potential links between identities and cognitive repertoires. That those links exist should be obvious. Our identities influence what we value, whom we know, and what we experience. They also influence how we make sense of those experiences.16
We live in segregated communities defined by identity categories, so those categories correlate with the knowledge, models, information, representations, and heuristics that constitute our repertoires. Our communities influence our opportunities and how we represent the world.17
A similar path of logic can explain differences between westerners and easterners. Easterners (and here I mean Asians, not New Yorkers) more often rely on relational representations, while westerners focus more on individual objects.18 An American will say, “Look at that fish.” A Japanese will say, “Notice the pattern of fish.” These differences derive from history, cultural practices, and experiences. They are not essential.
In thinking through the effects of identity on our cognitive repertoires, in places it will be helpful to distinguish between fluid intelligence and crystallized intelligence.19 Fluid intelligence corresponds to problem-solving skills and logical reasoning. Tests of fluid intelligence ask subjects to match patterns or solve logic puzzles. Tests of crystallized intelligence ask for the definition of cosine. A person with high fluid intelligence can acquire knowledge quickly. If he does not retain it, then he lacks crystallized intelligence.
Including this distinction complicates and enriches how we think about the components of a repertoire. Some parts of our cognitive repertoires, namely heuristics and representations, contribute to fluid intelligence. Other parts, like information, knowledge, and models, contribute to our crystallized intelligence.
It is worth noting that scores of fluid intelligence have risen over the past ninety years, and a majority of those gains occur among the lower half of the distribution.20 This rise in fluid intelligence may result from greater exposure to logical and abstract reasoning; that is, children learn heuristics and representations that enable them to score more highly on tests of fluid intelligence. Thus, we should think of fluid intelligence as dependent in part on life experiences.
How Identity Might Matter
The claim that members of identity groups have special understanding of their own groups can be supported by multiple strands of evidence. Data on friendships proves particularly convincing. People tend to hang out with people from the same identity group. One study finds that a typical white American has but a single black friend and a single Latino friend.21 It involves, therefore, a rather large logical leap to presume that most people possess deep knowledge of the preferences and beliefs of people with different identity classifications.22
Whether identity diversity plays a direct or indirect role, or no role at all, will depend on the situation. Often, we have no idea what type of diversity will come to bear. Entire books can be filled with anecdotes of these idiosyncratic diversity bonuses.
Here’s a favorite of mine, based on a physical difference told to me by Laszlo Bock, a former senior vice president of Google. After acquiring YouTube, Google found that approximately 10 percent of people uploaded their videos upside down. This was a puzzle. Well, it was only a puzzle to the 90 percent of people who are right-handed. When a left-handed Googler heard the problem, she knew the cause: lefties. A left-handed person tips her smartphone to the right to take a horizontal video. A right-handed person tips his smartphone to the left. What’s upside down to a righty is right side up to a lefty.
The smartphone anecdote sticks.23 It is simple, unexpected, concrete, credible, and uplifting. It also oversimplifies. It maps a single identity attribute or experience to a particular solution, masking the complex, mysterious relationship between who we are and how we think.
Such stories function better as motivational impetus than as guides to practice.24 Could Google have known that a left-handed person was needed? Should they make sure every programming and problem-solving team has a lefty? To build a logic of diversity bonuses from identity, we need to think through the mapping between identity bundles and cognitive repertoires. A good place to start that process is with (inappropriate) stereotypical mappings from identity characteristics to cognitive repertoires.
Stereotypical Identity-Based Cognitive Differences
Recall from the iceberg analogy that identity attributes—woman, Asian American, visually impaired—differ in their observability. We can make inferences based only on the attributes we see. We infer that identity diversity implies diversity of thought. We infer that Asian Americans think differently from European Americans, and that men think differently from women.
This assumes that we can take the parts of our cognitive repertoires and trace them back to individual components of our identities (see figure 4.3). We do this when we say, “We need a woman’s perspective on this,” or, “Our group includes no people of color. We’re missing out on valuable ways of thinking.” Though well intended, these statements lack precision (and border on offensive).
Unpacking the reasoning reveals the imprecision. Start with a population of people. Each person has multiple dimensions to her identity. Focusing on a single attribute puts her in a box with everyone in the population who shares that attribute. That could be the African American box. It could be the woman box. If the latter, the arrows oblige us to identify the parts of her cognitive repertoire that result from being female.
Let’s focus first on information and knowledge. A person’s information and knowledge includes everything she has acquired over her lifetime, what she has learned at school and in her vocation and avocations. Some of the information will be organized. Some will be idiosyncratic.
Figure 4.3 The (Oversimplified, Problematic) Single-Attribute Mapping
Some will depend on our identities. Identity-based categorizations—gender, race, religion, sexual orientation, and such—correlate with differences in books read, movies watched, websites visited, college courses taken, and so on by people who vary on that dimension.
Imagine convening a large group of women and men and asking everyone to list everything they know, all their information and knowledge. Each person would create a large set. We could label each set with an F or an M to denote gender. If a statistician compared those two collections of sets, she would find substantial overlap. Some information would be in almost every set. Almost everyone knows that cars run on gasoline, that New York City has tall buildings, and that horses have four legs.
The statistician would also find information and knowledge that reveals geographic location and educational background but provides no clue as to gender. Information on the capital of Kentucky and knowledge of mitosis and meiosis might well be equally present in the sets labeled F and M.
The statistician would also find bits of information and pieces of knowledge that do correlate with gender. She could identify statistical differences between the two sets. Those differences would allow her to evaluate someone’s set and predict the person’s gender. With enough data, she would classify correctly with a high degree of accuracy.
In my thought experiment, the statistician can see a person’s entire information and knowledge set. With far less information, algorithms can predict a person’s gender, age, and ethnicity. For instance, Google Ads’ settings page predicts your gender, age, and interests based on your search history.25 It pegs me as male.26 These algorithms exploit the fact that members of different identity groups differ statistically in the books they buy, the websites they visit, the movies and television shows they download, the health issues they research, and the sports and hobbies they pursue. These differences produce statistically distinct distributions over the websites searched by identity groups.
Men compose 60 percent of whiskey drinkers, 70 percent of baseball fans, 80 percent of private investigators, 90 percent of hunters, and 100 percent of vasectomy patients. A person who knows that Old Forester Classic is 86 proof, that Lou Whitaker was the 1978 American League Rookie of the Year, that it is easier to lift fingerprints from a golf ball than from a gun,27 that Washington and Idaho preclude the removal of the sex organs when gutting a deer so that the state can verify gender, and that residual pain from vasectomies can last more than a year would be statistically far more likely to be a man.
If the statistician had access to people’s full cognitive repertoires and again created two collections of sets, one labeled M and one labeled F, the two collections would again be internally diverse, overlap substantially, and exhibit enough statistical differences for the statistician to predict gender out of sample with a high degree of accuracy.
Similar results would hold if we considered classifications based on race. The cognitive repertoires of people labeled African American or Latino will be diverse, overlap with those of other racial groups, and have identifiable statistical signatures.
Subclusters also exist within each cluster. The people classified as Asian American include Japanese, Korean, Chinese, and Filipino Americans. Each of these groups possesses distinct histories and cultural practices. Variation within the set of Asian Americans may be clustered by country of origin. A Japanese American may have little knowledge of Korean American culture and vice versa.
This way of thinking of the connection between identity and cognitive diversity accords with the timber-framed house analogy. Second-generation Chicanas combine three identity groups: Chicano, female, and second-generation American. The cognitive repertoires of second-generation Chicanas consist of the intersection of the cognitive repertoires of those three groups.
That intersection property does not imply that a characteristic of each of those three groups is also a characteristic of Chicanas. It could be that Chicanos, women, and second-generation Americans disproportionately attend comedy clubs and know recent comedic tropes. If second-generation Chicanas do not go to comedy clubs, they would lack that knowledge. The failure of characteristics to apply to intersections is analogous to the problem identified by intersectionality theory. A firm can hire women and African Americans but not hire African American women. Here, a characteristic can be common among Chicanos, women, and second-generation Americans but not be common among second-generation Chicanas.
Given that within each identity category, people vary in their repertoires, we cannot expect token representatives to speak for an entire group. If the Ford Motor Company wants to market cars to the growing Latino population, it should hire a cohort of Latino employees, not a single Latino in marketing. Similarly, if the University of Delaware wants students with diverse cognitive repertoires to create a rich intellectual environment, they require more than one person from each identity group.
The cognitive repertoires of people within any identity category differ. No single person can represent her identity group. A cohort of people can function as a representative sample.
Can’t a Homogeneous Group Be Diverse?
The cloud analogy applies to any group with a common identity attribute—even the proverbial group of all white men. They too will be cognitively diverse. The information, knowledge, mental models, representations, and heuristics within their heads will differ. A group of older white men who attended Exeter and Dartmouth in the 1960s, majored in economics, and pursued careers in investment banking will be cognitively diverse: no two the same.
But they will not be that diverse. Their collective repertoire will contain holes. For example, the group may lack knowledge of the educational aspirations of second-generation Latinos or frameworks for explaining the cultural acclimation process of Somalian immigrants. Or they may bin all Asian American women within a single category, limiting their ability to predict the efficacy of many public health interventions.
These older white men would not be the first people we would turn to for heuristics for debugging Python code, writing hip-hop lyrics, or predicting trends in fashion. Nor would we expect them to be experts on women’s health concerns, to be familiar with the video game market, or to understand the financial concerns of recent widows.
I am purposefully stereotyping here. Some men may be experts in those domains. My point is that their clouds of information and knowledge, their clouds of heuristics and mental models, and their clouds of representations differ statistically from the clouds of twenty-five-year-old Latinas.
Whether it is then a large leap to infer that the lack of identity diversity on Wall Street contributed to the home mortgage crisis is an empirical question. It is probably fair to say that investment bankers had less knowledge of the financial models carried around in the heads of the diverse people flipping houses than they would have had if the investment bankers had been more diverse.
Figure 4.4 The Identity-Bundle-to-Cognitive-Repertoire Mapping
Whatever mapping does exist goes from identity bundles, from our whole selves, to our full repertoires (see figure 4.4). Each person’s unique bundled identity combines with experiences and training to produce her unique cognitive ensemble, situated somewhere in the cloud.
IDENTITY-BASED COGNITIVE DIFFERENCES BY TASK
The pragmatic question remains as to the amount to which identity differences in repertoires map to performance differences. Inferring the extent of identity’s influence is difficult because the mapping is complicated. If we represent people as timber-framed houses of identity attributes and represent identity groups as clouds, then, as already discussed, the concept of a woman’s perspective is not well defined. That term cleaves identity bundles and ignores the gestalt of repertoires. It isolates gender and representation. Applying similar logic, we should not put much stock in the notion of a Latino mental model, a heterosexual categorization, or an Asian American heuristic.
That said, empirical correlations between identity bundles and cognitive repertoires do exist. As mentioned, Google knows our identity attributes from our searches.
A first step in determining the relevance of the identity-driven components of our repertoires is to recognize that it varies by task. We expect the identities of city council members to influence their positions on educational reform more than we expect the identities of students to correlate with the techniques they use to invert matrices in linear algebra class.
Identity also interacts with geography. Each person has someplace she lays her head down on a pillow at night. If she also worships, works, organizes, and plays near that place, she will know the people who live there with a depth and granularity that outsiders cannot match. Given a seat at the table, she can add her information, knowledge, and mental models to policy choices. Following similar reasoning, health services providers who understand the lifestyles, community structures, and ways that information spreads within a community will be necessary parts of a health services team. We cannot just roll out a bunch of doctors with degrees from elite schools and expect them to improve health outcomes.
As a rule, we should expect identity-driven differences to matter in any domain that serves people: education, finance, entertainment, or health. Thus, a strong case can be made for the potential contributions of cognitive diversity correlated with identity diversity to the areas of product design and marketing, to policy creation, and to media production.
Identity diversity can also matter in less weighty social contexts. Careful thought and elaborate efforts enter into casting decisions. Interns, midlevel executives, and producers all read scripts and suggest actors. After someone has been selected, she may not accept the part. In the movie Gravity, Sandra Bullock stars as a lone astronaut who attempts to return to Earth in a failing space module. Angelina Jolie had been director Alfonso Cuarón’s first choice. Jolie opted not to pursue the project.
To find a replacement, Cuarón sought out other A-list Hollywood stars. He considered actresses he thought capable of carrying an entire film. His set included Natalie Portman, Blake Lively, and Scarlett Johansson.28 He made a wise choice in Bullock. The movie won six Academy Awards, and Cuarón was awarded the Oscar for Best Director.
Mitt Romney, the 2008 Republican candidate for president, described how he would use a similar process to populate a diverse cabinet by creating a binder of qualified women. Romney was pilloried for this in the media. That criticism is ironic given that creating lists of qualified women candidates guards against an implicit pro-male bias when making hires. Binders are a standard strategy for improving the representation of women.
To see why binders—or some systematic method for constituting a pool of candidates—are necessary because of our biases, I created an experiment I called Tom Hanks Is Busy. In the experiment, I describe the following scenario: Tom Hanks has been cast in a Hollywood drama as an everyman who rises to the occasion to rescue a group of teens in peril. For personal reasons, Hanks steps out of the role. The director needs to find a replacement. The experiment asks that you replace Tom Hanks.
Stop reading for a moment and think. Who would you choose to replace Tom Hanks?
I have posed this question to many people. Common answers include Liam Neeson, George Clooney, Adam Sandler, Hugh Grant, and Brad Pitt. These actors can be thought of as adjacent possibles to Tom Hanks. They are all white men.
To see how a systematic approach reduces bias, list some adjectives that describe Hanks. You probably chose terms like huge star, dependable, kind, sensitive, middle aged, funny, and artistic range. Now suppose I gave you that list of adjectives and asked you to think of actors with those attributes.
You might now think of Denzel Washington.
Is Denzel a good replacement for Hanks? Hanks was born in the mid-1950s and won two Academy Awards. Ditto for Denzel. Hanks has starred as a war hero and an everyman. Hanks played a pilot who crash-lands a plane in Sully, as did Denzel in Flight. Both are huge celebrities who star in movies that have grossed more than $2 billion. Both also direct and produce.
The have similar off-screen images. Both are quietly religious. Both have been married for more than twenty-five years and both have four children. Both donate to multiple charitable causes and both are active in politics. If Denzel did not jump to mind, that does not mean you are racist, but it does reveal that race plays a large role in the adjacent possibles. If you did not think of Sandra Bullock or Meryl Streep, then gender does as well.
This thought experiment should not be dismissed as lighthearted fun. The history of racial inequities in casting decisions within the film industry suggests casting directors rely on racially and gender-biased adjacent possibles. Identity-diverse casting teams would be one way to correct for those biases, that is, if African Americans would be more likely to think of Denzel.
A similar logic extends to strategic decisions for consumer products. A brewer wants to sell beer. Soft-drink companies want to sell soda. Knowing the occasions that bring people together and the foods that people eat informs marketing, packaging, and pricing. Paper companies want to sell diapers. In 2016, Pampers released a hilarious video of babies caught in the act, so to speak. In the initial release of the video, every baby was Caucasian. The majority of babies born in the United States are not.
Can Identity Diversity Matter for Science?
The relevance of identity diversity is less obvious on scientific and technical problems. We know that in some cases, our identities can have large effects. I begin with a well-known example in which identity played a role in scientific research.29
Fifty years ago, descriptions of the process of human egg fertilization characterized the sperm as conquering a passive egg. The people who wrote those descriptions were men. The film of attacking sperm that I watched in seventh-grade science class with a roomful of boys might have featured Richard Wagner’s Ride of the Valkyries as a soundtrack. In that version, the egg waits passively as the conquering sperm swarm. A more gendered analogy would be difficult to construct.
Later, a research team including both men and women showed that the egg actively selects from among the many sperm attacking its outer wall. They discovered that the egg’s selection depends in part on genetic diversity.30 Understandings of the sperm and the egg show how identity diversity can influence how people represent and model a phenomenon.
To dig deeper, let’s engage in a prototypical thought experiment regarding homogenous and diverse groups that goes as follows: a NASA administrator must select a group of scientists to increase the lift capacity of the new Space Launch System (SLS). She must choose between two groups. The members of both groups have similar ability as measured by IQ tests, college grades, and college entrance scores.
The first group consists of eight Muslim men between thirty-eight and forty-three years of age. All earned aerospace engineering degrees from Georgia Tech and have worked at the Marshall Space Flight Center in Huntsville, Alabama, for at least ten years.
The second group (the diverse group) consists of men and women belonging to a variety of ethnic and racial groups. They also vary in age, education, and work experience. Some earned aerospace engineering degrees. Others studied mechanical or electrical engineering. This group includes two outsiders on loan from Boeing and Ford. Should the administrator choose the homogenous group or the diverse group?
This experiment stacks the deck in favor of the identity-diverse group, whose members possess more diverse technical repertoires. Even if their identity diversity had no influence, the second group would be the stronger choice.
A more informative thought experiment varies identity diversity and training diversity independently. This requires four groups. The first would consist of people from the same identity group with similar experiences and training. A second group would include identity-diverse people with similar training and experiences. A third group would consist of people from the same identity group who have different educations, training, and experiences. The fourth group would be diverse in every possible way, like the second group from the previous example. We could then think through which group would perform best. That would be difficult if not impossible for any one person to do.
These types of thought experiments do not resemble real-world choice processes. A NASA administrator choosing a group of scientists to increase the lift capacity of the new SLS would not be given a binary choice between a “homogenous” group and a “diverse” group. Nor would she choose among four types of groups like those just described.
Instead, she would form a group from a set of applicants. In selecting that group, she would ask herself a series of questions including the following: What types of training will be relevant for the problems the team will face? What educational backgrounds might I consider? What experiences? Does the group include people with applicable knowledge bases? Does it include people who bring different frames—someone who will think of costs, someone who will think of risks, someone who will think of novel possibilities, and so on?
The answers to those questions will not be independent of identity diversity. Recall how the clouds overlap and differ statistically. Leaving out people from an identity group means selecting a nonrepresentative set of cognitive repertoires. If we have no evidence either way, should our default position be that those parts of the repertoire that correlate with identity matter or that they don’t?
To answer that question, we must think through the parts of a repertoire. The information and knowledge pertinent to designing a propulsion system might be thought to have the weakest correlation with identity. Most of that knowledge would come from engineering, though perhaps a piece of knowledge from biology, kinesiology, or chemistry could improve the design. Ideally, we would pull academic transcripts when forming the group to get a team with diverse knowledge.
Identity-based heuristics and mental models and frameworks surely matter more. When a problem becomes difficult, we seek analogies drawn from past experiences. Identity contributes to how we represent systems. One of the proposed designs for an SLS booster was named the Dark Knight.
Identity may also influence how we think about power and balance. Take the experience of learning to scull. Single sculling shells are over twenty-five feet long and less than eighteen inches wide. Those dimensions do not promote stability. Men are on average taller than women, and they have broader shoulders, which distributes more of their weight to the upper part of their bodies. The typical man placed in a shell struggles to balance. One good solution is to start rowing. Most women, who on average have a lower center of mass, can achieve balance by relaxing. Placed in the same situation, men achieve balance through power and women by, well, balancing. Those different experiences could translate into frameworks and heuristics for achieving balance when designing spacecraft and satellites.
Thus, a wise NASA director would consider identity diversity as contributing to cognitive diversity. To tap into as diverse a set of cognitive repertoires as possible, she would want teams with men and women. She would also want a variety of types of engineers who attended different schools. Standard engineering courses like advanced fluid dynamics may cover different topics at different schools. MIT covers windmills in its fluid dynamics course. The University of Illinois does not.
Though the direct influence of identity diversity on germane cognitive diversity in technical fields should be less pronounced than in domains that involve people, a lack of identity diversity may well correlate with less cognitive diversity. The NASA director’s mission, perhaps to get people to Mars and back, requires brilliant, diverse thinkers. She would be reluctant to put together a team of people who all belonged to the same identity group.
Furthermore, we must keep in mind that on unsolved problems, where we lack a heuristic or do not know what knowledge, model, or representation might lead to the breakthrough, we want as much relevant cognitive diversity as possible. Where a solution might come from, or where a person might acquire it, could be serendipitous. Add to the mystery of the source of breakthroughs the evidence that the presence of diverse others causes us to think differently, and we have even more reason to err on the side of identity diversity.
Identity-Based Opportunity Bonuses
Here, I tell three stories of how a unique way of looking at the world resulted in an innovation or a new product. In each case, the person recognized a dimension that others had overlooked. One recognized a medical disparity based on race. The other two saw unmet community needs. In each case, the dimension aligns with the person’s identity. All three people created diversity. I refer to these as opportunity bonuses because these diverse ways of seeing created new opportunities. They were not solving an existing problem.
Patricia Bath: Medical Innovator
At an early age, Patricia Bath displayed uncommon scientific talents. Her precocity as editor of the Charles Evans Hughes High School’s science paper led her to be invited to a 1959 National Science Foundation summer cancer research workshop held at Yeshiva University. While there, she derived a mathematical equation describing cancer cell growth. She was only sixteen years old.31
She went on to study chemistry and physics at Hunter College in New York and, at age twenty-six, earned an MD from Howard University. Bath later became the first African American resident in ophthalmology at New York University Hospital, the first African American woman surgeon at UCLA Medical Center, and the first woman to head a residency program at the Charles R. Drew School of Medicine and Science. She would later receive a medical patent for the Laserphaco Probe, a tool that removes cataracts while simultaneously irrigating the eye to facilitate surgical lens replacement.
At first glance, Patricia Bath’s story appears to be one of the millions of examples of the larger-pool logic. Had Bath not been invited to the summer workshop, had she not been allowed to study chemistry and physics and to earn an MD, the world would have been denied an accomplished surgeon and a medical innovator. Her success reminds us that talent knows no color or gender.
If we dig deeper, we find diversity bonuses. Bath worked at both Harlem Hospital and the Columbia University Eye Clinic Hospital. She held the former position partly because of her identity. Having these two windows on the world, she noticed a disparity in blindness rates among African Americans at the two hospitals. She was able to identify the lack of access to care as a cause. The research stemming from this observation led to the development of community ophthalmology. Her identification of the disparity—the differences in blindness rates—originated a new opportunity to improve health care that others did not see.
Robert Johnson: BET
Billionaire businessperson Robert L. Johnson was born in Hickory, Mississippi, on April 8, 1946, and raised in Freeport, Illinois. Johnson earned a BA in social science at the University of Illinois and an MA in public policy at Princeton. His work experience included stints as a public affairs director for the Corporation for Public Broadcasting, a communication director for the National Urban League, and a vice president of the National Cable and Television Association.
By 1980, he had built a unique and powerful repertoire of skills and knowledge: academic training in social science, an understanding of public and cable television markets, experience communicating through media to the African American community, and an awareness of the distinct viewing patterns of African Americans. In that year, he launched a cable channel, BET (Black Entertainment Television), that catered to the interests of African Americans.
Beginning with a two-hour block of programming on Nickelodeon, Johnson offered content ignored by the white media. He noticed that the music video channel MTV relied on a largely British collection of extant videos. Other than Tina Turner, few black artists received airtime on MTV. In a 1983 on-air interview, David Bowie observed, “It occurred to me having watched MTV over the last few months that it’s a solid enterprise. It’s got a lot going for it. I’m just floored by the fact that there’s so few black artists featured on it. Why is that?”32
Johnson was already two steps ahead. He had launched Video Soul, a half-hour show featuring R&B and soul music. Johnson also produced Black College Football, broadcasts of games between all-black powerhouses like Jackson State and Grambling. Though these schools sent as many players to the NFL as traditional schools like the University of Southern California, Ohio State, and Florida, their games received no airtime on major networks. Johnson rectified that and launched a media empire that propelled him to riches.
Johnson became the first African American billionaire and the first black majority owner of a major sports franchise. He succeeded because he knew music and sports, because he had developed strong analytic skills, because he had a range of experience in television and broadcasting, and because he had developed tools to communicate to black audiences. He had the right repertoire.
Johnson’s story embodies how a person’s identity and experiences can reveal an opportunity that others miss. Emerson wrote that “people only see what they are prepared to see,” a sentiment echoed by Canadian playwright Robertson Davies, who wrote that we only see “what the mind is prepared to comprehend.” Johnson comprehended a large, untapped market of African American viewers and accumulated the skills to leverage that opportunity.
Christy Haubegger: Latina Magazine
Christy Haubegger provides another example of how an identity-based perspective created a new market. Christy, a Mexican American, grew up in Bellaire, Texas, as the adopted daughter of “tall blond people.” After graduating from the University of Texas, she completed her law degree at Stanford in 1992, where she edited the Stanford Law Review.
Law degree in hand, Haubegger looked around and saw no role models from her identity group. She had no Anthony Kennedy or Sandra Day O’Connor whose footsteps she could follow. Sensing an opportunity, she founded Latina magazine. She would create her own role models by profiling Latina astronauts, authors, judges, and entrepreneurs.
Today, Latina’s circulation exceeds a quarter million and has garnered many honors, including being named best magazine by Advertising Age. The stories of Johnson and Haubegger demonstrate how a person’s identity can influence the opportunities that one sees. It is not coincidental that Johnson started BET or that Haubegger founded Latina magazine. The opportunity bonus they produced stemmed in part from their identities.
Interactions and Identity Diversity
Up to now, I have considered the direct effect of identity diversity on cognitive repertoires. Identity diversity can also indirectly influence how people tap into their cognitive repertoires. Evidence from controlled experiments shows that the presence of a person from a different identity group causes people not from that identity group to generate more ideas and construct more complex arguments.33
As a thought experiment, imagine a group of people evaluating designs for a public garden. Add to this group a person in a wheelchair. That person will almost surely cause his fellow group members to pay attention to lane widths and curb heights. Add a blind person to the group and others will be more aware of physical boundary markers. Differently abled people contribute to the cognitive diversity of the group without saying a word.34
The presence of diverse others can also lead to more critical thinking. In a simulated stock market experiment, traders were assigned to either homogeneous or ethnically diverse groups. The members of diverse groups questioned price deviations more critically and produced fewer, and smaller, bubbles. An analysis shows that their market prices were nearly 60 percent more accurate.35
The logic of diversity bonuses demonstrates how diverse cognitive repertoires contribute to better outcomes on a variety of tasks. As we think about what causes people to differ in what they know, how they represent problems, the models and frameworks they apply, the categories they use, and the techniques they master, we cannot but conclude that education and life experiences play major roles.
As we contemplate that question more deeply, we realize that our identities play a significant role on many problems. They influence the categories we construct. They influence the dimensions on which we focus. When analyzing a topic like economic inequality, women are more likely to consider gender effects and members of minority groups are more likely to bring up racial disparities. They do so because those dimensions are salient.
To some extent, our identities influence how each of us goes about any task. We bring to it our unique standpoints. We also apply ways of thinking we learned in school and technical skills we learned on the job. We cannot separate out the beams of the timber-framed house, nor can we necessarily assign how a person thinks about a problem to either her identity, her experiences, or her education. On some problems, they may all matter.
Last, as we think about the influence of identity diversity, we cannot forget that our identities also influence what we value and deem worthy of our time and attention. The decision about what problems we address, which is itself a complex question, therefore requires an identity-diverse team. We must be inclusive in deciding our goals if we want to create effective inclusive groups and teams to achieve them.