Lack of diversity creates blind spot
Facebook’s mostly white and male vetters failed to stop fake Russian ads
SAN FRANCISCO — In a heated moment during last week’s Capitol Hill hearings, Rep. Terri Sewell questioned whether the paucity of African Americans in Facebook’s workforce — and more specifically on the teams of reviewers who vet content and advertising — contributed to the company’s failure to catch Russian operatives using fake accounts to stoke racial tensions ahead of last year’s presidential election.
Displayed behind Sewell, a Democrat from Alabama and a member of the Congressional Black Caucus, was one of the Russian-backed ads showing a famous black-and-white photograph of the Black Panthers from 1968 with the caption: “Never forget that the Black Panthers, group formed to protect black people from the KKK, was dismantled by us govt but the KKK exists today.”
The Facebook ad was intended to exploit racial divisions and get AfricanAmerican users to follow a fake Russian account called Blacktivist. It was shared on Facebook at least 29,000 times.
“Who are your vetters, and are they a diverse group of people?” Sewell confronted Facebook’s general counsel Colin Stretch.
“Like every aspect of our workforce, we are committed to building a workforce that is as diverse as the community we serve,” Stretch replied.
“You’re saying I should trust that your vetters, who will be vetting this kind of information, are going to be a diverse workforce?” Sewell asked.
“What you should be confident of is that we understand the importance of diversity,” Stretch said.
From Washington to Silicon Valley, that confidence is sorely lacking. Despite repeated pledges to close the racial gap in its U.S. workforce, a tiny fraction — 3% — of Facebook is African-American. In all, Facebook employs 259 black people, according to the company’s most recent government filing. That’s out of 11,241 people.
The Russian infiltration of Facebook is just the most recent example of how that lack of diversity creates major blind spots for the giant social network that’s staffed mostly by white and Asian men, says University of Southern California professor Safiya Umoja Noble, author of the upcoming book
Algorithms of Oppression: How Search Engines Reinforce Racism.
The reality: With people screening content on Facebook spread around the globe, many of them working as lowly paid contractors, it’s possible that few, if any, are African American and therefore are far less familiar with and invested in racially coded messages about the Black Lives Matter movement and civil rights issues in the U.S., Noble says.
“We have to ask how it is that con- gressional members have enough knowledge to recognize the problematic nature of the kind of hateful and divisive advertising that is allowed to proliferate on these platforms, but that some of the best engineers in Silicon Valley, and their CEOs, can’t acknowledge the significance of it,” she said.
Sewell’s face-off with Facebook during Wednesday’s House Intelligence Committee hearing on Capitol Hill follows weeks of criticism from the Congressional Black Caucus over Facebook’s handling of fake Russian pages and ads.
The pages and the ads they placed, rigged to look like the work of American activists, spread incendiary messages during and after the presidential campaign. Facebook repeatedly denied the Russians exploited its platform until September. Last week, the company admitted that 146 million Americans may have been reached on Facebook and Instagram.
The black community had raised suspicions about those posing as activists on Facebook and Twitter.
When the Russian-backed “Blackti-
Facebook employs 259 black people, according to the company’s most recent government filing. That’s out of 11,241 people.
vist” Facebook page and Twitter account called for a march in Baltimore after the police custody death of Freddie Gray, Rev. Heber Brown III, pastor of a Baltimore church, confronted Blacktivist, asking if those behind the account organizing the police brutality march were local. The account responded that it was not based in Baltimore but was “looking for friendship, because we are fighting for the same reasons.”
Brown retorted that Blacktivist should “come learn and listen before you lead.” When he later learned that the account was fake and based in Russia, Brown said he was stunned that he had disrupted a “Russian op.”
Another fake Facebook page operated by the Internet Research Agency in St. Petersburg, Russia, Black Matters, promoted a protest in New York City the Saturday after the election with a Facebook ad. More than 16,700 people signed up to attend on the event page and 33,000 said they were interested in the event. Tens of thousands showed up.
That such an extensive covert effort by an adversarial foreign power mimicked U.S. political discourse and gained big followings by targeting specific groups of American Facebook users and spreading racially divisive messages is deeply worrying to Allison Jones, director of communications of Code2040, a San Francisco organization that works to increase the diversity of the tech industry’s workforce.
Jones says igniting racial animosity is a common technique used for centuries to exploit divisions to amass and maintain power. Now, with the enormous reach and influence of social media in today’s society, it’s more effective than ever, making it even more crucial that Facebook and other technology companies employ diverse workforces, she says.
“These technologies are increasingly the filter through which we interpret the world around us. There is huge potential for them to increase empathy, understanding and connectivity, but there are also great risks that they will only amplify biases already held, deepening inequity as well as polarizing people across the country,” Jones says.
The Congressional Black Caucus is moving with greater urgency to pressure Silicon Valley to hire more people of color.
Last month, representatives met with Facebook’s Chief Operating Officer Sheryl Sandberg in Washington, then with executives in Facebook’s Menlo Park, Calif., headquarters. They are speaking up about the lack of diversity at all levels of the company, from the board of directors to the engineers who build the company’s products. Sandberg has committed to recruiting an African American to the all-white board. Even fewer African Americans — just 1% — hold technical roles at Facebook.
“Having a diverse workforce in all aspects of tech companies, especially as relates to reviewing content and ensuring user safety, is critical to the success of the tech industry,” Rep. G.K. Butterfield, a North Carolina Democrat and former chairman of the caucus, told USA TODAY in an email.
Having more black people in the room is a crucial step but isn’t enough, says Sara Wachter-Boettcher, author of Technically Wrong: Sexist Apps, Biased Algorithms, and Other Threats of Toxic Tech.
African Americans are far more attuned than white Americans to the ways racially divisive messages are coded and communicated, but companies should not place an unfair burden on “the most marginalized staff to speak up and push back against company decisions,” she says.
“This is a topic where actual experts exist — people with backgrounds in critical race theory, in history, in cultural studies,” she says. “Why aren’t policies and review procedures also informed by people who’ve studied the way racism has historically functioned in America, and how racist groups communicate and gain power?”
Facebook’s chief executive Mark Zuckerberg said he plans to double the number of employees and contractors who handle safety and security issues to 20,000 by the end of 2018. He said Facebook would pour more resources into engineering efforts and would focus on new artificial intelligence systems to detect what Zuckerberg described as “bad content and bad actors.”
Noble says Facebook is placing too much faith in yet-to-be-proved artificial intelligence to monitor content on its platforms. These technologies are not yet sophisticated enough to detect the political and social nuances of the carefully crafted manipulation deployed by the Russians.
And, too often, she says, they harbor the biases of the people who build them and the biases they learn from the Internet.
When artificial intelligence expert Rob Speer, chief science officer of Luminoso, built a restaurant review algorithm, it rated Mexican dining spots lower because it learned to associate “Mexican” with negative words like “illegal.” The reason, he says, is that the system learned the word “Mexican” from scouring the Web, where Mexican is frequently used with the word “illegal,” as in “illegal immigrants.”
A Stanford study found that Internettrained artificial intelligence associated stereotypically white names with positive words such as love and laughter and stereotypically black names with negative words such as failure and cancer.
Having too few people of color building and testing new products has created public relations nightmares for Silicon Valley companies.
In 2015, Google apologized for its photos application mistakenly labeling photos of black people as “gorillas.” The Google workforce is 56% white and 2% black.
In 2016, Google image searches were criticized for delivering very different results for “three black teenagers” than for “three white teenagers,” raising troubling questions about how racial bias in society and the media is reflected online.
“For some time, we have heard claims by Silicon Valley executives who are in denial about the ways their projects impact democracy and vulnerable people — from racist and sexist trolling to fraudulent disinformation that poses as credible evidence we can trust,” Noble says. “In reality, these companies generate all kinds of ideas and unleash them on society, and we have to deal with the repercussions and harm after the fact.”