Good morning, Broadsheet readers! Beto O’Rourke has a surprising set of supporters in Texas, a #HimToo meme takes an unexpected and encouraging turn, and Amazon struggles with gender bias in AI. Have a terrific Thursday.
• Alexa, eliminate algorithmic bias. With the on-going AI revolution has come plenty of discussion about the risks posed by artificial intelligence and its machine-learning capabilities. Chief among them is the worry that AI will reinforce biases by relying on data that reflects the world’s current marginalization of women and people of color.
While this discussion has often been theoretical, Amazon has now given us a prime example of what that kind of discrimination looks like in real terms.
According to Reuters, the e-commerce giant killed an AI-powered recruiting tool it spent years developing because its engineers couldn’t figure out how to stop the system from discriminating against female job candidates. Amazon wanted the system to evaluate resumes and identify top prospects. To achieve this, Amazon fed the system a decade’s worth of CVs from candidates who’d applied to jobs at Amazon—most of whom were men. As a result, the system taught itself to downgrade resumes that included the word “women’s” and applicants who’d graduated from two women’s-only colleges. Amazon’s team tried to correct the bias, but ultimately concluded it was impossible to stop the tool from finding new ways to discriminate against female candidates. Reuters reports that the company’s HR department used the system to generate candidate recommendations but never relied on it entirely when filtering for possible hires.
The episode gives life to technologists’ greatest fears about AI and reinforces the need to heed warnings issued in this new age.
As Amazon’s own Toni Reid, who leads the Alexa unit, told the Fortune MPW Summit crowd last week: “Be thoughtful about who you’re bringing to the table when you’re building these products.”
Likewise, Aicha Evans, chief strategy officer of Intel, advised against underestimating humans’ own computing capabilities. “Life is about figuring out purpose, connecting dots, making judgments,” she said. “Computers don’t know how to do that yet.”
But even as technology breeds algorithmic bias, technology may also hold the answer to fixing—or at least revealing—it. Microsoft, for instance, announced this spring that it’s developing a way to spot bias in artificial intelligence algorithms with the goal of allowing businesses to rely on the technology without discriminating against certain groups of people. IBM launched a tool in September to analyze how and why algorithms make decisions in real time. And Facebook has tested a system for flagging racial, gender, or age biases in machine-learning algorithms. And what was its first application, interestingly enough? Facebook’s jobs algorithm that matches job seekers with firms looking to hire.
ALSO IN THE HEADLINES
• Designing an exit. Architect Richard Meier was accused of sexual harassment by five women seven months ago. Now, Meier is finally stepping down as head of the firm Richard Meier & Partners, but he will still “remain available to colleagues and clients who seek his vast experience and counsel.” Meier is 83, and the consequences sound like a semi-retirement. Curbed
• Beto’s base. There’s a surprising reason challenger Beto O’Rourke, a Democrat, has had such a strong campaign against Texas Sen. Ted Cruz. Some white evangelical women, long a safe bet for Republicans, are switching sides to vote for O’Rourke. “I care as much about babies at the border as I do about babies in the womb,” one voter said. New York Times
• A year after Ronan. To mark the one-year anniversary of its own story about Harvey Weinstein, The New Yorker had its essayists across literature, food, culture, and news devote their space to the #MeToo movement. All offer compelling perspectives on underreported aspects of #MeToo: the legacy of black women’s testimonies, punishment for individual abusers versus systemic justice, the language of abuse and victimhood, harassment in the restaurant industry, the movement’s reception across generations, and how #MeToo has eluded government surveillance in China. The New Yorker
• Mic drop. We all know of mansplaining and manterrupting—but how about manhandling the mic? In a debate between state representative candidates in Minnesota, Jamie Mahlberg experienced the behavior first hand, when her opponent, Rep. Duane Quam, grabbed the microphone while she was still answering a question. When he finished talking, he tried to get her to take the mic back; when she wouldn’t, he dropped it in front of her. The break in debate decorum and invasion of personal space says a lot about gender dynamics in bids for office. Washington Post
MOVERS AND SHAKERS: Former CIA deputy director of intelligence Jami Miscik joins the board of GM. Square’s CFO Sarah Friar will step down and take a job as CEO of Nextdoor. Lindsay Peoples Wagner of The Cut is the new editor of Teen Vogue. At Hearst, sweeping staffing changes: Jessica Pels replaces Michele Promaulayko as editor of Cosmopolitan, Joanna Saltz takes over for Sophie Donelson as editorial director of House Beautiful, and Kristin Koch will be executive director of Seventeen.com, replacing Joey Bartolomeo. Steele Marcoux will be editor-in-chief of Veranda.
IN CASE YOU MISSED IT
• On track? Formula One has a tough glass ceiling: no woman has raced since 1976. In an effort to diversify the sport, a new all-women competition, W Series, will debut in 2019 with a $1.5 million prize. Some drivers, however, say the competition will set women back by segregating them from other competitors. CNBC
• Coffee care. Starbucks has a new benefit for its baristas: 10 days’ worth of subsidized care providers for children or adults. Starbucks employees can pay $1 an hour for in-home care for their family members or $5 a day for a daycare center through Care.com. This is the last addition to the rollout of improved benefits the company announced in January. CNN
• Meme mom. If you weren’t tuned into the meme-cycle this week, I suggest you catch up on this one. It started when a Twitter user posted a photo of her son in a Navy uniform with the hashtag #HimToo and said he wouldn’t “go on solo dates” because of a fear of false sexual assault allegations. The Twitterverse quickly turned her post into a meme, with people posting photos of their own “sons.” But then things took an unexpected and heartwarming turn: The son posted his own version of the lampooned photo, said that he believes women, and that “sometimes the people we love do things that hurt us without realizing it.” You can follow that guy Pieter Hanson at @thatwasmymom. Vox
• WORTH works. A new program at a prison in Connecticut is trying to make sure that women between 18 and 25 don’t end up back behind bars. Therapy, treatment for addiction, mentorship with older women who are also serving time, and interview and resume training are all ways the program, Women Overcoming Recidivism Through Hard Work, is fighting recidivism. The Marshall Project
ON MY RADAR
Who does ‘Southern femininity’ really belong to? Vox
Under their eye: The rise of Handmaid’s Tale-inspired protesters Vanity Fair
How marketers talk about motherhood behind closed doors The Atlantic