CES Panel Examines Problem of Bias in Artificial Intelligence

As artificial intelligence is increasingly embedded into devices and experiences, the problem of racial and gender bias has become apparent, in several embarrassing and disturbing incidents. The industry has paid attention to studying how bias is introduced — often via the underlying data — and how to fix it. Former FCC commissioner Mignon Clyburn, now at MLC Strategies, led a discussion with Helloalice.com president Elizabeth Gore and Uber head of inclusive engagement Bernard Coleman on the topic.

Clyburn, who revealed that, “one report predicted that, by the end of this year, spending in [AI] will eclipse $46 million,” asked Gore and Coleman how their companies are working to overcome bias. Helloalice.com started as a platform for women entrepreneurs, said Gore, and is now open to men, but still prioritizes “women, people of color, those with disabilities, U.S. military vets and the LGBTQ community.”

“We built Alice on the premise that we wanted to provide best business services to the new majority of business owners,” said Gore. “It gives you advice, not just based on your physical location and where you are in your career, but your gender, if you have a disability and your race among other factors. We want to mine as many opportunities as we can.”

She gave the example of Alice co-founder Carolyn Rodz, a Latina with dual passports. “What pops up on her screen in addition to normal opportunities are Latina business circles, peer-to-peer mentoring for Latinas, and similar offerings.” Clyburn summed it up as “customization as a brand benefit.”

Coleman, who has been at Uber for three years, noted that “inclusive engagement makes sure the workforce is representative of society.” “How we show up in different markets is important,” he said. Gore noted that women get less than two percent of venture capital money, and women of color get a fraction of one percent.

“Do women even get to walk in the room,” she asked. “Alice is working to make sure access to capital is getting to the people most qualified — and that means all of us!”

Creating diversity in AI is not a “build it and they will come” proposition. “We committed to 50 percent diversity,” said Gore. “But we had to go to the organizations that work with those populations to introduce the technology to them and that meant a lot of partnerships.”

Clyburn asked if it’s possible to build an unbiased AI if it’s impossible to have an unbiased human. Coleman pointed out that people who start a company bring in their friends, who look like them. “When we work in a vacuum, we haven’t included others to get more data,” he said. “It’s an incomplete dataset. But we have to manage expectations what AI can and can’t do. Part of the bias problem is giving it too much power.”

“It’s better to start with values when you’re small, because it’s very expensive to catch up,” said Gore. “People, planet and profit is the way of the future and expected by consumers. It’s better for business and the bottom line.”