ACM Interactions

In a “buddy movie,” two people confront a problem. One is often calm and analytic, the other impulsive and intuitive. Initially distrustful, they eventually bond and succeed by drawing on their different talents.

This captures the core elements of a case for diversity: When people with different approaches overcome a natural distrust, their combined skills can solve difficult problems. They must first learn to communicate and understand one another. In addition to the analyst and the live wire, buddy movies have explored ethnically, racially, and gender diverse pairs, intellectual differences (Rain Man), ethical opposites (Jody Foster’s upright agent Clarice Starling teamed with psychopathic Hannibal Lecter), and alliances between humans and animals or extra-terrestrials.

Diverse buddies are not limited to duos (Seven Samurai, Ocean’s Eleven). All initially confront trust issues. Lack of trust can block diversity benefits in real life, too: Robert Putnam demonstrated that social capital is greater when diversity is low and that cultures with high social capital often fare better. However, the United States has prospered with high immigration-fueled diversity, despite the tensions. When is diversity worth the price?

In the movies, combining different perspectives solves a problem that no individual could. The moral case for racial, gender, social, or species diversity is secondary, although these differences may correlate with diverse views and skills. At its core, diversity is about survival, whether the threat is economic failure or the Wicked Witch of the West. “Don’t put all your eggs in one basket,” financial planners advise. A caveat is that diversity is not always good. Noah had to bar Tyrannosaurus rex from the ark; it wouldn’t have worked out. For a given task, some of us will be as useful as the proverbial one-legged man at an ass-kicking party. Exhortations in support of diversity rarely address this.

Diversity in teams receives the most attention. My ultimate focus is on the complex task of managing diversity in large organizations—companies, research granting agencies, and academic fields. But a discussion of diversity and survival has a natural starting point.

Biological diversity

Diversity can enable a species to survive or thrive despite changes in environmental conditions. Galapagos finches differ in beak size. Big-beaked finches can crack tough seeds, small-beaked finches ferret out nutritious fare. Drought or a change in competition can rapidly shift the dominant beak size within a single species. The finches do well to produce some of each.

In two situations, biological diversity disappears: A species with a prolonged absence of environmental challenge adapts fully to its niche, and a species under prolonged high stress jettisons anything nonessential. If circumstances shift, the resulting lack of diversity can result in extinction. In human affairs too, complacency and paranoia are enemies of diversity.

Human diversity

Coming to grips with workplace diversity is difficult because all forms come into play. Our differences span a nature-nurture continuum. Race and gender lie at one end, acquired skills at the other. Shyness or a preference for spatial reasoning may be inherited; cultural perspectives are acquired. In his book The Difference, Scott Page focuses on the benefits of diverse cognitive and social skills in problem-solving. As I wrote, a friend announced a startup for which a major investor was on board provided that other investors join: He wants diverse concurring reviews.

A team matter-of-factly recruiting core skills doesn’t think of them as diversity—but an unusual skill becomes a diversity play. Whether differences originate in nature or nurture isn’t important. Understanding their range and how they can clash or contribute is.

Drawing on thousands of measurements and interviews, William Herbert Sheldon’s 1954 Atlas of Men [1] yielded three physical types, each focused on an anatomical system accompanied by a psychological disposition: (i) thin, cerebral, ectomorphs (central nervous system), (ii) stocky, energetic mesomorphs (musculature), and (iii) emotional, pleasure-seeking endomorphs (autonomic nervous system). Consider a team comprising a scarecrow, lion, and tin man keen to establish ownership of a brain, courage, and a heart, or Madagascar’s giraffe, lion, and hippopotamus.

Aldous Huxley used Sheldon’s trichotomy in novels. Organizational psychologists favor broader typologies. Companies know that good teams can be diverse and try to get a handle on it. A popular tool is the Myer-Briggs Type Indicator. This 2x2 typology was built on Carl Jung’s dimensions of introversion/extraversion and thinking/feeling. It is consistent with his view that a typology is simply a categorization that serves a purpose. Other typologies are also used. Early in my career, my fellow software developers and I were given a profiling survey that I quickly saw would indicate whether we were primarily motivated by (i) money, (ii) power, (iii) security, (iv) helping others, or (v) interesting tasks. If I filled it out straightforwardly I’d be (v) followed by (iv). Instead, I drew on my childhood Monopoly player persona, and in the years that followed received very good raises. I took the initiative to find interesting tasks, rather than relying on management for that.

Teams and organizations

Organizations generally differ from teams in several relevant respects. Organizations are larger, more complex, and last longer [2]. An organization, a team of teams, requires a greater range of skills than any one team. Organizations strive to minimize the time spent problem-solving, where diversity helps the most, and maximize the time spent in routine execution. Most teams continually solve problems; one change in personnel or external dependency can alter the dynamics and lead to a sudden or gradual shift in roles.

Should an organization group people with the same skill or form heterogeneous teams? Should a company developing a range of products have central UX, software development, and test teams, or should it form product teams in which each type is represented? Homogeneous teams are easier to manage—assessing diverse accomplishments is a challenge for a team manager. Diverse teams must spend time and effort learning to communicate and trust.

Homogeneous teams could be optimal for an organization that is performing like clockwork, heterogeneous teams better positioned to respond in periods of flux. A centralized UX group is fine for occasional consulting, an embedded UX professional better for dynamic readjustment.

Teams

Consider a working group with a single manager, such as a program committee for a small conference, an NSF review panel, or a team in a tech company. The scope of work is relatively clear. Diversity may be limited: quant enthusiasts may keep out qualitative approaches or vice versa; a developer-turned-manager may feel that a developer with some UX flair has sufficient UX expertise for the project.

Where does diversity help or hinder? Joseph McGrath identified four modes of team activity: taking on a new task, conflict resolution, problem-solving, and execution. Diversity often slows task initiation. It can create conflicts. Diversity is neutral in execution mode [3], where a routine job has been broken down into component tasks, minimizing complex interdependencies.

Scott Page describes contributions of diversity to the remaining mode of team activity, problem-solving. Diversity helps when clearly recognizable steps toward a solution can be taken by any team member, as in open source projects or when several writers work on dialogue for a drama. Although a team executing in unison like a rowing crew may not benefit from diversity, most teams encounter problems at times. Members are often collocated, enabling informal interaction, learning to communicate, and building trust. When resource limitations force hard decisions on a team, members understand the tradeoffs. Subjective considerations sometimes override objective decision-making on behalf of team cohesion: “We just rejected one of her borderline submissions, let’s accept this one.” “His grant proposal is poor but his lab is productive, let’s accept it at a reduced funding level.” In contrast, responses to organizational decisions are often less nuanced.

Teams have teething pains, conflicts, and managers who can’t evaluate workers who have different skill sets or personalities. But in general, diverse teams succeed. One that fails is replaced or its functions reassigned.

When time is limited, introducing diversity is challenging. I was on a review panel that brought together organizational scientists and mathematicians. The concept of basic research in organizational science mystified the mathematicians, to whom it seemed axiomatic that research on organizations was applied. Another review panel merged social scientists studying collaboration technology and distributed AI researchers; the latter insisted loudly that every grant dollar must go to them because DARPA had cut them off and they had mouths to feed.

Organizations

Organizations often endorse diversity, perhaps to promote trust in groups that span race, gender, and ethnicity. However, it is rarely a priority. Given that managing diversity is a challenge, why should a successful organization take on more than necessary? An organization’s long stretches of routine execution don’t benefit from a reservoir of diversity that enhances problem-solving. Complacency sets in. Perhaps diversity would yield better problem-solving, outweighing the management costs. Perhaps not.

The biology analogy suggests that a reservoir of skills could enable an organization to survive an unexpected threat. We don’t need big beaks now, but keep a few around lest a drought appear. A successful organization outlives a team, but few surpass the 70-year human lifespan. Perhaps identifying and managing the diverse skills that could address a wide-enough range of threats is impossible; managing the clearly relevant functions is difficult enough.

One approach is to push the social, cognitive, and motivational diversity that aids problem-solving down to individual teams to acquire and manage, using tools such as employee profile surveys. Unfortunately, it doesn’t suffice for organizational purposes to have skills resident in teams. Finding and recruiting a specific skill that exists somewhere in a large organization is a nightmare. I have participated in several expertise-location system-building efforts over 30 years, managing two myself. The systems were built but not used. Incentives to participate are typically insufficient. Similarly, cross-group task forces have been regarded as stop-gap efforts that complicate normal functioning.

Another approach is to assign teams to pursue diverse goals. For example, one group could pursue low-risk short-term activities as another engages in low-probability high-payoff efforts, drawing on different skills or capabilities.

Assessment at large scale

Organizations can’t be infinitely diverse. A company does a market segmentation and narrows its focus. NSF balances its investment across established and unproven research. A conference determines a scope. When unexpected changes present novel problems, will a reserve of accessible skills and flexibility exist? Management can draft aspirational mission statements, but in the end, responsiveness is determined by review processes, such as employee performance evaluations, grant funding, and conference and journal reviewing.

A pattern appears as we scale up: Assessing across a broad range not only requires us to compare apples and oranges (and many other fruits), it requires deciding which apples are better than which oranges. Sometimes all the apples or all the oranges are discarded.

Large organizations. How broadly should assessments and rewards be calibrated laterally across an organization? Giving units autonomy to allocate rewards can lead to the perception or fact that low-performing units are rewarded equivalently to stronger units. A concerted effort to calibrate broadly takes time and can lead to the dominant apples squeezing out other produce. For example, when rewards are calibrated across software engineering, test, and UX, the more numerous software engineers to whom “high-performing UX professional” is an oxymoron can control the outcome.

Organizations also sacrifice diversity to channel resources to combat exaggerated external threats. A hypothetical company with consumer and enterprise sales could respond to a perceived threat to its consumer business by eliminating enterprise development jobs and devoting all resources to consumer for a few years. When the pendulum swings back to enterprise, useful skills are gone.

Granting agencies. An agency that supports many programs has three primary goals: (i) identify and support good work within each program (a team activity), (ii) eliminate outdated programs, which facilitates (iii) initiating new programs, expanding diversity. Secondary diversity goals are geographic, education outreach, underrepresented groups, and industry collaboration.

Let’s generously assume that individual programs surmount team-level challenges and support diversity. The second goal, eliminating established programs that are not delivering, can be close to impossible. Once a program survives a provisional introductory period, it is tasked to promote the good work in its area—there is an implicit assumption that there is good work. Researchers in a sketchy program circle the wagons: They volunteer for review panels and for rotating management positions, submit many proposals (“proposal pressure” is a key success metric), rate one another’s proposals highly, and after internal debates emerge with consensus in review panels.

The inability to eliminate non-productive programs impedes the ability to add useful diversity. For example, NSF has a process for new initiatives that largely depends on Congress increasing its budget. The infrequent choices can be whimsical, such as the short-lived “Science of Design” and “CreativIT” efforts [4]. I participated in three high-level reviews in different agencies where everyone seemed to agree that science suffers from inadequate publication of negative results, yet we could find no path to this significant diversification given current incentive structures.

Large selective conferences. Selective conferences in mature fields form groups to review papers in each specialized area. Antagonism can emerge within a group over methods or toward novel but unpolished work, but the main scourge of diversity is competition for slots, which causes each group to gravitate toward mainstream work in its area. Work that bridges topic areas suffers. Complete novelty finds advocates nowhere. Researchers often wistfully report that their “boring paper” was accepted but their interesting paper was rejected.

Startups: Team and organization

A startup needs a range of skills. It may avoid diversity in personality traits: People motivated by security or power are poor bets. Goals are clear and rewards are shared; there is a loose division of labor with everyone pitching in to solve problems. The short planning horizon and dynamically changing environment resemble a team more than an established enterprise. With no shortage of problems, diversity in problem-solving skills is useful, but every hire is strategic and there is little time to develop trust and overcome communication barriers.

Professional disciplines

Competition for limited resources works against community expansion and diversity. Two remarkably successful interdisciplinary programs, Neuroscience and Cognitive Science, originated in copious sustained funding from the Sloan Foundation. In contrast, I’ve invested more fruitless time and energy than I like to think about trying to form umbrella efforts to converge fields that logically overlapped: CHI and Human Factors (1980s), CHI and COIS (1980s), CSCW and MIS (1980s), CSCW and COOCS/GROUP (1990s), CHI and Information Systems (2000s), and CHI and iConferences (2010s). An analysis of why these failed appears elsewhere.

Conclusion

This is not the short essay I expected, and it doesn’t cover the equity considerations that drive diversity discussions in university admissions and workplace hiring. What can we conclude? Noting that diversity requires up-front planning to possibly address unknown future contingencies, I will consider where the biology analogy does and doesn’t hold.

With moderate uncertainty, diversity is a good survival strategy; with major resource competition, diversity yields to a focus on the essentials. So avoid exaggerating threats. Next, given that choices are necessary, what dimensions of diversity should we favor? Ecological cycles favor the retention of capabilities that were once useful—a drought may return. Economic pendulum swings argue for the same.

However, the march of science and technology creates both obsolescence and novel opportunities and challenges. Some, but not all, can be anticipated by studying trends. It is fairly empty to recommend focusing on efficiency in execution while retaining flexibility, but “avoid overreacting to perceived threats” is again good advice. Businesses narrow when they should diversify. Government funding is poured into defense and intelligence at the expense of health, education, infrastructure, and environment. And finally, my favorite hot button example, large conferences.

HCI researchers have always been terrified of appearing softer than traditional computer science and engineering. So we followed their lead. We drove down conference acceptance rates, kept out Design, and chased out practitioners. But other CS fields evolve more slowly, with greater consensus on key problems. Human interaction with computers explodes in all directions. Novelty is inevitable, yet with acceptance rates of 15% - 25%, each existing subfield accepts research central to today’s status quo, leaving little room for research that spans areas, is out of fashion but likely to return, involves leading edge practitioners, or is otherwise novel [5]. Could our process consign us to be followers, not leaders?

Endnotes

1. His atlas of women was not completed.

2. To be clear about my terminology use, a “football team” is an organization, although the group of players on the field together is a team. Boeing called the thousands of people working on the 777 a team, but here it would be an organization.

3. An exception is an organization tasked with problem-solving. The World War II codebreaking organization at Bletchley Park made extraordinary use of diversity, documented in Sinclair McKay’s The Secret Lives of Codebreakers: The Men and Women Who Cracked the Enigma Code at Bletchley Park (Plume, 2010).

4. DARPA is an agency with top-down management which can and does eliminate programs, sometimes restarting them years later.

5. See Donald Campbell’s provocative 1969 essay “Ethnocentrism of Disciplines and the Fish-scale Model of Omniscience.”

Thanks to Gayna Williams for ideas and perspectives, John King, Tom Erickson, and Clayton Lewis for comments on an earlier draft.

Jonathan Grudin