Facebook’s global director of diversity Maxine Williams recounted at the Wharton People Analytics Conference how in a previous position with a global law firm, she spent two years living in Geneva. She submitted the necessary paperwork to rent an apartment there. After she had moved in, the cleaning person for the previous tenant knocked on the door to see if the new renter wanted to continue the service.

“Oh, I didn’t expect to see you here,” said the cleaning person, sounding surprised.

The woman didn’t know Williams personally, and upon questioning it turned out she was referring to Williams’ dark skin. She told Williams matter-of-factly, “The owner said she would never rent to black people,” adding that the embassy of Congo had recently tried to rent the apartment without success.

Sure enough, said Williams, the next day the white owner came to her door to tell her she had to move out. Williams asked if it was because she was black. The owner indignantly denied it and said she had simply decided she didn’t want to rent out the apartment any more.

“My white in-laws were aghast: ‘How could they do that?’” said Williams. Her response was that things like this happen all the time, but almost never does the person being discriminated against get the information outright. Williams said the cleaning person had in a way “given her a gift” and that she was so pleased at the time, she had hugged her.

“Any time you interact with somebody who’s under-represented, get over yourself and your own identity, and think about theirs.” –Maxine Williams

For Williams, the experience with the owner illustrated “second-generation bias”: subtle racism that many people are subjected to in their business and personal lives. Discrimination is no longer open, accepted and even sanctioned by law as it once was: “We’re not any longer sitting at a lunch counter where somebody can say ‘we’re not going to serve you.’” But as a result, the fact that you are being treated differently can be hard to put your finger on.

“This is the awful thing: It leaves you feeling like you could be crazy,” said Williams.

In the workforce, she noted, this bias can manifest itself when a person in an under-represented group sees “a whole bunch of people who look like you at the bottom levels, and people who don’t look like you at the top.” Could it really just be chalked up to missed promotions? Some academics who study this phenomenon call it “something in the water,” said Williams. She made the point that generally in America, at any level, job function, or company (with the exception of minority-owned businesses), there is under-representation particularly of black and Hispanic people.

Williams stated that in her career, social justice has been her focus. As an attorney she fought for workers’ rights and civil rights both in her native Trinidad and in the U.K. Williams has also collaborated with several international organizations on human rights issues. But she considers her most important credential for addressing bias in the workplace to be the fact that she is black and a woman. “There’s some stuff you [just] know,” she stated. She added that when she speaks to under-represented groups about unconscious bias, “everybody knows what I’m talking about.”

How Race Affects Professional Advancement

As a diversity professional, Williams was excited by the prospect of people analytics when it first came on the scene. She recapped that the goal of people analytics is to move away from gut decisions and prioritize action based on data. “We might abbreviate it to ‘evidence-based talent management,’” she said. “It sounded fabulous.”

She recognized its enormous potential for reducing unconscious bias and creating opportunity for under-represented groups: to help make progress on “problems that human beings have not found solutions for.”

But there was a catch, said Williams. She explained that when studying any population statistically, a researcher needs a large enough sample size in order to draw confident conclusions. And when it comes to trying to make data-informed talent management decisions about very under-represented people in the workforce such as blacks and Hispanics, Williams found she was consistently told by data experts, “Sorry, but the N is too small.” (N refers to population size in statistics.)

Williams commented, “So it’s almost like they’re saying to me, ‘Hmm, if there were more of you, we could tell you why there are so few of you.’ Now who sounds crazy, right?” She added that she had spoken with a number of diversity directors at other companies and found that they were all grappling with the same issue.

Some hard evidence does exist about how race influences promotion at work. Williams cited eye-opening statistics from the Ascend Foundation, an organization focused on Asian advancement in the workplace. She said a 2014 study comparing data from five technology companies found that white men and women were 150% more likely to be promoted to the executive level than their Asian counterparts. The study also found that while gender, too, had a dampening effect on professional advancement, race had an impact that was more than 3.5 times greater than gender.

Williams noted that at Facebook, building diversity is not regarded as an “add-on” or “nice to have” but essential to its business success. Related to racial and ethnic diversity is cognitive diversity. She explained, “We’re very focused on how do we get as many people as possible on as many teams as possible with different ways of seeing the world — differences in information and skills and background — because we have complex problems to solve.”

Williams was asked how managers, especially majority-group managers such as white men, can be of help to co-workers from under-represented groups. She answered that there are two basic things that under-represented people need.

“The reality of marginalization is the invisible thumb on the scales. We have the opportunity to build the right scales to address things as they really are, not as we pretend they are.” –Maxine Williams

“They need to know that they are not crazy,” she said, referring to the unconscious bias that they suspect they are being subjected to. “And they need to know that they will be supported.”

She acknowledged that such conversations can be difficult to initiate, but advised, “Any time you interact with somebody who’s under-represented, get over yourself and your own identity, and think about theirs.” Tell them you might not have the answer but you are a willing, able, and supportive colleague, she said.

She exhorted the audience of people analytics professionals to think about developing new benchmarks and processes that will serve under-represented groups. “The reality of marginalization is the invisible thumb on the scales,” she said. “We have the opportunity to build the right scales to address things as they really are, not as we pretend they are.”

Are We Hard-coding Bias into the System?

The promise and limitations of today’s people analytics were also discussed by panelists in the session “An Analytical Perspective on Diversity.”

Corinne Low, a Wharton professor of business economics and public policy, talked about how using people analytics data can help us avoid stereotypes. As an example, she described how most people visualize a potentially successful entrepreneur.

“We have this idea that the next startup founder has to be that young guy in a hoodie who’s been poking around in his basement since he was 10 years old,” said Low. But the data says otherwise. Low cited a recent study from University of California Berkeley’s Haas School of Business which found that the average age of a successful startup founder is 38.

She pointed out, though, that as human resource professionals try to create objective measures for judging candidates, they may unwittingly be hard-coding in biases from the past that actually hinder diversity. For example, an HR professional at a software company might learn that engineering candidates who were involved in software or startup clubs in high school have a high rate of success. They might implement a resume review system programmed to give extra points to incoming resumes that list these clubs.

“But what if there were no women in the startup club in high school?” said Low. That doesn’t necessarily mean that women wouldn’t be successful in these jobs, she said, but it does mean they could get left behind in the hiring process.

Derek Avery, a senior associate dean for diversity and global initiatives at Wake Forest University School of Business, agreed that people analytics has a “garbage in, garbage out” problem. He observed that professionals need to be careful when trying to build algorithms that encourage diversity, because currently there isn’t even agreement in the field about exactly what diversity is.

Professionals need to be careful when trying to build algorithms that encourage diversity, because currently there isn’t even agreement in the field about exactly what diversity is.

He referred to the famous quotation from Supreme Court justice Potter Stewart on the definition of pornography: “I know it when I see it.” The problem with applying this to discrimination, said Avery, is that often we don’t know it when we see it.

Joelle Emerson, founder and CEO of Paradigm, a strategy firm that helps organizations become more inclusive, talked about some innovative ways companies can reach beyond traditional HR channels to build diversity. She described a recent project with a social media company focused on improving its referral pool so it was not so homogenous.

Emerson’s remedy was to have a senior-level executive issue an email to the whole company stating the problem and asking employees to spend time over the next six weeks “thinking about your own network … thinking about who might add diversity to this team, and refer them.” She reported that over that six-week period, the company saw a 30% increase in the number of women referred, and a 5,500% increase in the number of blacks and Hispanics referred.

Emerson explained that sometimes when employees consider who to refer to their company, they may be unconsciously trying to “match” what the existing workforce looks like. But “actually having people think about their [personal] networks a little bit differently can have a huge impact,” she said.

Another recent study Emerson described involved having a C-suite team keep track of, over a month, the amount of time they spent in informal conversations with people on their teams. This might include exchanging a few words in the hall, chatting over coffee, or having lunch. Upon seeing the results, she said, “They were all shocked by the homogeneity of the people they were having these little informal chats with and what that added up to.”

Emerson said that one proposed solution was to have the executives schedule more one-on-one meetings with people who weren’t reaping the benefits of these casual, but potentially important, conversations.