Search Engine Bias
You’ve probably heard the riddle about the boy who, after surviving a car accident that killed his father, arrives at an emergency room in need of surgery. The surgeon says “I cannot operate on this boy because he is my son.” How is this possible? The surgeon is the boy’s mother, but researchers at Boston University found that only 15% of children age 7 – 17 could correctly solve this riddle, cause sexism.
Gender stereotypes for professionals are reinforced in many ways, including searches in popular search engines. As Marian Wright Edelman, founder of the Children’s Defense Fund, once said, “You can’t be what you can’t see.” According to a Pew research study, 57% of searches for professionals under-represent women. Learn more about the image problem in search engines.
Try searching the word “professor” in Google images, and you’ll be hard-pressed to find photos that include non-white and non-male scholars. It’s even worse when you search for illustrations of professors. Here at The Representation Project, we found this out the hard way when planning an impromptu office celebration to congratulate our Executive Director, Dr. Caroline Heldman, on becoming a full professor at Occidental College. When we searched images of professors (hoping to sketch a doodle on her card), we found a slew of images of little Einsteins—which one can only assume Google believes to be the only possibility for a professor. Sexism has found its way into machine learning and search engine algorithms—and unfortunately for us, society’s unconscious bias is mirrored in search engine results.
By design, people trust the highest ranking search engine results. However, search engines have programmed in preference for white men and gender stereotypes when searching for visual representations of many members of our society. When searching for images of “schoolgirls” in Google—results don’t return pictures of young women in classrooms, libraries, or gathering on a school campus as students tend to do. Instead, the screen is flooded with images of women in sexualized, Halloween costumes.
Try searching for images of a “CEO” or a “film director.” While the results may paint an unfortunately accurate picture of the state of gender representation in those two professions, they also serve as a harmful authority— one that shows the person searching for images what society believes to be the norm for a CEO, film director, and countless other professions.
Gender bias is coded into search algorithms. With so few women and people of color in lead roles in tech, white men have a disproportionate say in the final results. Women make up only 11 percent of executive positions in Silicon Valley, while women of color are virtually nonexistent in leadership roles.
It’s obvious the search giants lack perspective when it comes to diversity. Search engines rank men significantly higher than women, with women being more associated with “kitchen” than “CFO” in Google’s world. Fortunately, stock image services like Shutterstock and iStockPhoto fared much better in our sample “Professor” search—returning real-world results showing a diverse range of women and men at the lectern and showing what’s possible when it comes to accurate representation in image searches.
Take action! Install the Google Chrome extension S.H.E. (Search Human Equalizer) for more balanced gender representation in search. You can even submit searches to the extension so that it can learn what women already know.