The importance of global marketsThe global markets that emerging technologies open to today’s businesses are extremely promising. Denise Pirrotti Hummel,[i] CEO of cross-cultural advisory firm Universal Consensus, points out: “Our economic development will forever be defined as our ability to succeed internationally. PwC forecasts India’s real annual GDP growth until 2050 at 8.9 percent, Vietnam’s at 8.8 percent, and China’s at 5.9 percent. The list of fast-growing emerging markets goes on and on. The U.S. forecast is a meager 2.4 percent, comparable with most Western economies. The domestic companies that are likely to see incremental growth in the coming decades are those that are not only doing business internationally, but that are developing the strategic skill set to master doing business across cultures. Cross-cultural core competence is at the crux of today’s sustainable competitive advantage.”
Impediments to cross-cultural awarenessThe problem, however, with mastering cross-cultural awareness lies in how immersed individuals typically are in their own culture. This is true even in our local workplace. We rarely recognize the biases we have toward our own worldview because they are deeply ingrained in us. We trust them because we have formed them over years of experience. As a result, when faced with perspectives that don’t match our own, we tend to interpret them as flawed and dismiss them. Such cultural inflexibility is deadly not only to expansion into markets with unfamiliar cultures, but also to the ability to innovate. Maintaining a diverse mix of perspectives has repeatedly proven to be essential for innovation. A 2016 Harvard Business Review article[ii] related a study of how homogeneity of perspectives can hinder group interactions. “In the experiment, teams were asked to solve a murder mystery. First, students were individually given 20 minutes to study the clues and pinpoint the likely suspect. Next, they were placed into teams of three with fellow members from the same [fraternity or sorority] house and given 20 minutes to discuss the case together and provide a joint answer. Five minutes into the discussion, however, they were joined by a fourth team member, someone from either their own house or another one. After collectively naming their suspect, members individually rated aspects of the discussion. More diverse groups — those joined by someone from outside their own fraternity or sorority — judged the team interactions to be less effective than did groups joined by insiders. They were also less confident in their final decisions. Intuitively, this makes sense: On a homogenous team, people readily understand each other and collaboration flows smoothly, giving the sensation of progress. Dealing with outsiders causes friction, which feels counterproductive. But in this case their judgments were starkly wrong. Among groups where all three original members didn’t already know the correct answer, adding an outsider versus an insider actually doubled their chance of arriving at the correct solution, from 29% to 60%. The work felt harder, but the outcomes were better. In fact, working on diverse teams produces better outcomes precisely because it’s harder.” Studies discussed in another Harvard Business Review article[iii] supported the same concept. In one series of studies, some financially literate individuals were grouped in homogenous groups and others in diverse groups and asked to price stocks in simulated markets. The diverse groups were 58% more likely to price stocks correctly than the homogenous ones. And, in another study of 4,277 existing companies, researchers found a direct correlation between the level of diversity on the companies’ research and development teams and those companies’ introduction of radical innovations over a two-year period. Add to that McKinsey research[iv] that shows that gender-diverse companies are 15% more likely to outperform the national financial median for their industry and ethnically diverse companies are 35% more likely and it’s clear that diversity aids innovation. Homogenous groups are quicker to reach consensus because their similarities lead them to consider fewer details. Diverse groups consider more details and make a more comprehensive analysis.
Bias and AIAs if cultural inflexibility was not enough of an impediment to successful globalization, yet another problem affects AI. Human biases are often unintentionally transmitted into AI systems by their developers. That’s because the data that AI systems use to conduct analyses is not something that naturally occurs in our world and that developers “capture.” Instead, it is something that developers create. Someone chooses what is included in a dataset and what is excluded. That’s where the problem arises. If the data that is chosen to be included is chosen with an unintended bias, the dataset will contain that bias and its conclusions will reflect that bias. Hannah Wallach,[v] a senior researcher in Microsoft’s New York research lab, points out: “We often talk about datasets as if they are these well-defined things with clear boundaries, but the reality is that as machine learning becomes more prevalent in society, datasets are increasingly taken from real-world scenarios, such as social processes, that don’t have clear boundaries.” The problem goes deeper than just the researcher’s unrecognized biases. Data that originates from information that has been affected by historical inequities can carry those historical biases into the dataset and affect the conclusions reached. Kate Crawford, a senior researcher in that same lab, adds: “The people who are collecting the datasets decide that, “Oh this represents what men and women do, or this represents all human actions or human faces.” These are types of decisions that are made when we create what are called datasets. What is interesting about training datasets is that they will always bear the marks of history, that history will be human, and it will always have the same kind of frailties and biases that humans have.” This process of choosing what data is included in a dataset and what is not is what led to the incident of the AI system that went against all medical knowledge and recommended that asthma patients suffering from pneumonia not be hospitalized. The data on which the system’s recommendations were based came from records that showed which pneumonia patients were chosen for hospitalization over the course of doctors’ treatment of their pneumonia. It didn’t include doctors’ actions that were taken at the first sign of pneumonia, only what occurred later in treatment. Thus, no asthma patients were admitted later in treatment; they had all been admitted at the first sign of symptoms. Because the records of initial actions were not part of the dataset from which the system drew, the AI system saw no asthma patients admitted in the dataset it was given. Thus, it wrongly concluded that asthma patients should not be prioritized for hospitalization. The same type of improper conclusion drawn from skewed datasets has been detected in AI systems designed to help judges determine which convicted felons should be incarcerated and which could be given probation. The datasets used were based on past sentencing patterns. As a result, the recommendations mirrored past practices of incarcerating a disproportionately large number of black felons while giving probation to a greater percentage of white ones. This shows that developers must consider not only the data itself, but also the context from which that data arose to avoid building unwanted biases into their systems. Brad Smith and Harry Shum, in their Microsoft whitepaper, The Future Computed,[vi] point out: “[B]ecause AI systems are designed by human beings and the systems are trained using data that reflects the imperfect world in which we live, AI can operate unfairly without careful planning. To ensure that fairness is the foundation for solutions using this new technology, it’s imperative that developers understand how bias can be introduced into AI systems and how it can affect AI-based recommendations.”
Diversity as a safeguard against unintended biasesDiversity in development teams serves a second purpose beyond providing more comprehensive analysis to AI system design. It also brings a greater range of perceptions of the data that can help the team avoid building unintended biases into the system. It also brings a greater range of perspectives into team discussion of what the problems are that the system needs to address. Research shows that more diverse teams choose more diverse problems to work on. This leads to more innovative solutions. Timmit Gebru, a post-doctoral researcher at Microsoft’s New York lab, says: “In some types of scientific disciplines, it doesn’t matter who finds the truth, there is just a particular truth to be found. AI is not exactly like that. We define what kinds of problems we want to solve as researchers. If we don’t have diversity in our set of researchers, we are at risk of solving a narrow set of problems that a few homogeneous groups of people think are important, and we are at risk of not addressing the problems that are faced by many people in the world.“[vii] An example of this can be clearly seen in a mistake that PwC narrowly avoided making[viii] in its early efforts at increasing diversity. A group of men were tasked with reversing a trend in which PwC’s workforce was disproportionally low on women, despite an aggressive effort to hire women straight out of college. The group’s initial assumption was that the women who left did so to start a family. The men thought the problem could be solved by PwC developing more programs to support new mothers. But further data-driven analysis showed that those who left were doing so from entry level positions – a time before women typically left the workforce to start a family. At all other levels, a larger percentage of men left than women. But the aggressive effort to hire women for entry-level positions was not being carried into hiring for positions that required more experience. There, the replacements hired were overwhelmingly men. The problem was not that women starting a family were finding PwC unsupportive. It was a breakdown in diversity hiring at more experienced levels. Had the group gone with its initial assumptions, their solution would not have addressed the real issue. This typifies again why homogenous groups are less effective than diverse groups. Their shared experiences and viewpoints can lead to premature conclusions that overlook the real issue. Fortunately, PwC’s practice of data-driven decision making ultimately saved the group from pursuing a plan that would have been a costly dead-end. Development of AI systems, as much as possible, should reflect the demographic of people who will be affected by the recommendations that those systems make. Smith and Shum point out: “[I]ndustry and academia should continue the promising work underway to develop analytical techniques to detect and address potential unfairness, like methods that systematically assess the data used to train AI systems for appropriate representativeness and document information about its origins and characteristics.” That is one of the key issues that groups like the AI Now Institute[ix] seeks to address. The Institute brings together people from a wide range of disciplines and backgrounds to study the implications that AI has for our future, so it can most effectively avoid unintended biases and truly represent the full spectrum of people who will be affected by it.
Incorporating diverse perspectives throughout the organizationThat is not to say that such work should be left to such cross-discipline, academic groups. It should be practiced within individual businesses, as well, if those businesses are to avoid unintended bias.
Rethink how you measure diversityHow you measure progress toward greater diversity is crucial to attaining it. Most organizations that set their sights on increasing diversity base their measurements on organization-wide metrics. Those can be deceptive, though, as they were for PwC before it started analyzing on a more granular level. What appears to be a diverse mix of backgrounds and demographics across an organization can be misleading if targeted groups predominantly fill lower-level positions and higher levels remain largely homogenous. Diversity consultant Aubrey Blanche[x] encourages organizations to measure diversity by job function and even by team to get a true picture of progress. This is especially crucial in teams responsible for innovation and expanding penetration into culturally diverse new markets.
Rethink where you target diversity initiativesSimilarly, make sure you’re applying diversity targets to all levels of your organization, not just to entry-level positions. Are you striving to add diverse perspectives when hiring for higher-level positions? Are upper-level candidates measured strictly on demonstrated skills and experience, or is the absence of more prestigious job titles that typically accompany those skills and experience that often dogs minority candidates counted against them? Striving to bring diverse perspectives into all levels of an organization doesn’t stop with hiring. Are promotions and access to “special projects” whose importance and visibility put participants in those projects on a fast track for advancement assigned disproportionately to those who mirror the homogenous nature of higher levels of management, or do all demographics and backgrounds have an equal opportunity to climb the ladder? Do minority employees advance in their careers in your organization at an equal pace to non-minority employees, or is there a pattern of them having to work twice as hard to get half as far half as fast? Diversity activist Bärí A. Williams[xi] points out: “Diversity gets people into the room, but inclusion keeps them there. True diversity is about more than just numbers; it must come with a heavy dose of inclusion. That means a company must be intentional about creating and fostering a culture where everyone has a seat at the table, not just entry to the room to watch as a bystander.” Inclusiveness is what is essential to bringing the benefits of a wide range of perspectives to spur innovation in organizations.
Rethink expectations about diversityIncorporating diverse perspectives into teams goes beyond merely hiring people from different backgrounds. Recognize that friction will be part of the process. Don’t expect to build a homogenous group of diverse people. Placing one person from a different background in an otherwise homogenous group does not promote diversity. In practice, it can lead the group to grant wary tolerance to the “outsider” and the “outsider” to feel pressure to “fit in” with the culture of the homogenous group to have their opinions heard. Recall the results of the Harvard Business Review article with the fraternity and sorority groups. The groups felt less confident of their decisions even though they achieved superior results. This was because they had to work harder to find the solution when presented with more comprehensive data. But find the best solution they did. The same unease will affect diverse groups in business – even as they function more effectively. Working in a diverse team feels harder but is more effective. Negative perceptions caused by being forced to step outside our comfort zone to consider other viewpoints provides a comfortable excuse to resist diversity efforts and retreat to our comfort zone. This expectation of an unhealthy level of conflict in diverse groups often leads managers to resist or water down diversity. It can even lead managers to abandon the process as being troublesome and ineffective. But that’s just part of our implicit biases. The benefits of applying diverse points of view to solving problems or forming strategies far outweighs the unease that examining additional points of view creates. The scientific evidence is clear, and it is unanimous: Working in diverse groups encourages more comprehensive analysis and leads to superior results. Achieving diversity involves far more than just entry-level hiring practices. It needs to become part of the DNA of not only the whole organization, but of individual teams. When pursued at this comprehensive level, it pays benefits far beyond what can be imagined.
Rethink the use of global “perks”Many global professional services firms have adopted global mobility programs where employees get to spend some time in other countries, or even relocate there. These programs have been treated more as a perk than as training, but the interactions that those engaged in such programs experience aids in broadening their perspectives by exposing them to unfamiliar cultures. Peter Lacy,[xii] managing director of strategy for the Asia-Pacific region at Accenture points out: “Our clients increasingly operate seamlessly across borders. Our people need to be able to do the same. That mindset comes from being exposed to new business cultures and experiences that come with international placements.” “Exposed to new business cultures.” A foreign placement isn’t about introducing Head Office perspectives to overseas operations – in fact, it’s the reverse. It’s about bringing the perspectives of overseas operations back to Head Office. Qatar is a stunningly wealthy country. Would you like to grow your business there? How are you going to do that if you don’t understand how Qataris think, what they value, what they disapprove of and what they really want from you? Because their demands are unlikely to be the same as those of your customers in Omaha. Employees with global experience are more attuned to recognizing the differing viewpoints of others and adapting strategies to successfully work with them. Participating in diverse teams, working abroad, etc., should be a personal performance metric. Many companies still measure the benefits of their global mobility programs merely through staff satisfaction and retention. The scientific research that supports the idea that diversity leads to better business performance, however, suggests we should budget more for such programs, involve more personnel in them and figure out how to calculate the benefits they bring to the company’s bottom line.
Rethink hiring processesA growing number of companies, including Apple, have taken steps to eliminate unintended bias from creeping into the resume review process, where decisions are made on who is invited for interview (and who isn’t). They now use blind recruiting applications that hide names, photos and dates. A growing number of other major companies are taking an opposite approach. They intentionally seek minority applicants through minority career sites like Jopwell to find qualified black, Hispanic and Native American candidates. Another approach was developed by global tech consulting firm ThoughtWorks. Their interview process[xiii] relies on teams who, after interviewing a candidate, immediately debrief each other to detect what they call “bad smells” in each other’s impressions. These “bad smells,” though, are not negatives in the candidate, but in the interviewers. They are the negative gut reactions that undetected biases often create. By examining vague negative feelings such as, “he didn’t feel like a good fit” or “she talked too much” that, in most companies, would exclude the candidate from further consideration, interviewers can better determine which impressions are qualification-based and which are bias-based. Another strategy that has proven effective is to stop telling people to be more inclusive and start involving them in the process of growing more diverse teams. Research shows that involving people in diversity taskforces[xiv] that approach the issue in the same no-excuses way that the business approaches profitability helps break down implicit biases. The more a person works to enhance diversity, the more that person comes to recognize its value.
TakeawaysGranted, the topic of diversity produces eyerolls from many employees. PwC CEO Dennis Nally, in a 2013 opinion piece,[xv] gave reasons why the word diversity itself has become divisive: “When we talk about diversity globally, word choice can impede progress. In mature markets, people are often tired of talking about it – they have ‘diversity fatigue.’ In emerging markets, they don’t always believe diversity is relevant, because they see it as an ‘imported’ concept.” The issue is not about trying to achieve someone’s idea of fairness or political correctness. It’s about bringing in new and fresh perspectives to spur innovation and develop more effective strategies as markets increasingly globalize. Incorporating more diverse perspectives has a measurable benefit to organizations’ bottom line and should be pursued as an aid to profitability.
[i] Denise Pirrotti Hummel, Understanding the Importance of Culture in Global Business, Oracle, May 2012, Available: http://www.oracle.com/us/corporate/profit/archives/opinion/050312-dhummel-1614961.html [ii] David Rock, Heidi Grant and Jacqui Grey, Diverse Teams Feel Less Comfortable – and That’s Why They Perform Better, Harvard Business Review, November 22, 2016, Available: https://hbr.org/2016/09/diverse-teams-feel-less-comfortable-and-thats-why-they-perform-better [iii] David Rock and Heidi Grant, Why Diverse Teams Are Smarter, Harvard Business Review, November 4, 2016, Available: https://hbr.org/2016/11/why-diverse-teams-are-smarter&_redirected [iv] Vivian Hunt, Dennis Layton, Sara Pierce, Why Diversity Matters, McKinsey&Company, January 2015, Available: https://www.mckinsey.com/business-functions/organization/our-insights/why-diversity-matters [v] John Roach, Debugging data: Microsoft researchers look at ways to train AI systems to reflect the real world, Microsoft, December 4, 2017, Available: https://blogs.microsoft.com/ai/debugging-data-microsoft-researchers-look-ways-train-ai-systems-reflect-real-world/ [vi] Brad Smith and Harry Shum, The Future Computed, p.61, Microsoft, 2018, Available: https://news.microsoft.com/futurecomputed/ [vii] John Roach, Debugging data: Microsoft researchers look at ways to train AI systems to reflect the real world, Microsoft, December 4, 2017, Available: https://blogs.microsoft.com/ai/debugging-data-microsoft-researchers-look-ways-train-ai-systems-reflect-real-world/ [viii] The PwC Diversity Journey, PwC, September 2016, p. 8, Available: https://www.pwc.com/gx/en/diversity-inclusion/best-practices/assets/the-pwc-diversity-journey.pdf [ix] AI Now, a research institute examining the social implications of artificial intelligence, AI Now Institute, Available: https://ainowinstitute.org/ [x] Aubrey Blanche, Tech Firms Striving for Diversity Fixate on the Wrong Metric, Wired, April 5, 2017, Available: https://www.wired.com/2017/04/tech-firms-striving-diversity-fixate-wrong-metric/ [xi] Bärí A. Williams, 8 Ways to Measure Diversity That Have Nothing To Do With Hiring, Fortune, April 20, 2017, Available: http://fortune.com/2017/04/20/workplace-diversity/ [xii] Janina Conboye, How valuable is international work experience?, Financial Times, November 6, 2013, Available: https://www.ft.com/content/89b6ebca-3a35-11e3-9243-00144feab7de [xiii] Emily Peck, This Company Proves You Can Hire More Women In Tech Right Now. No More Excuses!, Huffington Post, June 18, 2015, Available: https://www.huffingtonpost.com/2015/06/18/heres-how-you-get-more-wo_n_7613670.html [xiv] Frank Dobbin and Alexandra Kaley, Why Diversity Management Backfires (And How Firms Can Make it Work), Harvard University, February 26, 2015, Available: https://ethics.harvard.edu/blog/why-diversity-management-backfires-and-how-firms-can-make-it-work [xv] Dennis Nally, Stop talking about diversity, PwC, March 7, 2013, Available: http://pwc.blogs.com/ceoinsights/2013/03/stop-talking-about-diversity.html