If you do an online search for the greatest scientists of all time, you’ll likely find Albert Einstein and Stephen Hawking near the top. But it may take some digging to get to Rosalind Franklin, a chemist whose research helped reveal the components of DNA.
The gender biases that permeate the broader society are also reflected online, where there are 3.5 billion searches a day. But a new computer program aims to weed out such disparities.
Users of the Google Chrome web browser can install S.H.E., which will remix search results to ensure women are more realistically represented, according to Pantene, the hair care products company, which assembled a team of data scientists and tech engineers to create the new program.
“Women came to us saying it’s unfair that when we search for the greatest writer in history almost no women show up,” says Ilaria Resta, vice president of Procter & Gamble Hair Care, which owns Pantene, noting that’s one of several examples. “We felt we needed to do something … I have a daughter, 11 years old, and they do research at school with Google or other search engines and what I want my girl to see is an equitable world.”
S.H.E., which stands for “Search. Human. Equalizer,” basically works behind the scenes so that when people click on certain professions, results for women at the top of their fields will pop up higher instead of appearing intermittently or on pages searchers rarely click on.
Google Chrome warning: Beware! Google Chrome address bar can reportedly be used to launch a phishing attack
Amazon ups its delivery game: Amazon looking to shift to one-day Prime shipping and delivering orders wherever you want
Will comic book shops get an Avengers bump?: Have local comic book shops benefited from decade of Marvel films?
The tool will attempt to root out other forms of bias as well, such as how terms like “school girl” often bring up eroticized images, “teacher” fills the screen with pictures of women, and how a phrase like “man is to computer programmer as woman is to” will auto-populate “homemaker” in the last part of the sentence.
It’s not that computer algorithms are automatically biased, says Resta. But they adjust and “learn” based on how the public engages with them. If users, operating from their particular perspectives or preferences, tend to tag, caption or share only images of male engineers, the algorithm will tend to over-represent men versus women when that search term “engineer” is typed in.
A study published in December by the Pew Research Center found certain professions showed a particular disparity. Women showed up in 10% of searches for CEO, for instance, though 28% of chief executives are female, according to 2017 data from the U.S. Bureau of Labor Statistics.
Women also often show up further down in the search results, roughly four pictures lower than the first result on average, according to the Pew study. That matters because many people online don’t delve very far, sticking primarily to the first page of results.
The S.H.E. tool does not aim to skew the results. “What we want to avoid with S.H.E. is moving from one bias to another,” Resta says. “We’d like to be as data based as possible….’’
The hope is that increased awareness of how biases can affect search results and a tool to help correct it will bring more balance to future searches as users become more mindful of what they click on and share.
“It’s really to be a conversation starter and allow people to be more savvy when they look at the results.” Resta said, noting other areas that need to be addressed may be the results for families or couples which tend to emphasize the traditional nuclear idea or heterosexual pairings rather than the fuller array of both in society. “I expect people to frankly come to us if there are other biases we’re not aware of.”