68% of people say tech creators inject their own biases in products, study says
A recent study asked if participants believe that technology contains biases. The majority’s answer: a resounding yes.
The study, conducted for the Fortune 2018 Brainstorm Tech conference, showed that 63% of respondents felt that tech does more to justify biased decisions instead of removing bias from the decision-making process. Of the 2,100 people who were surveyed, 70% felt that tech amplifies people’s biases instead of diminishing them. More than two-thirds of the people surveyed blamed technology creators for injecting their own biases, instead of blaming the tech itself.
According to Fortune, Jon Cohen, the head of research for SurveyMonkey, said the study was a first of its kind, citing that the words “technology” and “bias” do not appear together in another major survey.
Bias in tech products manifests in various ways. Google made headlines back in 2009, when the search term “white people stole my car” automatically suggested directly below “Did you mean: black people stole my car.”
Facial recognition is susceptible to bias, too. Facial detection software is created by testing on actual faces. If a facial recognition app isn’t trained using an adequate amount dark-skinned faces, it will have a harder time parsing those types of faces. With law enforcement using facial recognition to locate criminals, a mix up could have dire consequences.
We see similar bias surface in the realm of virtual assistants like Apple’s Siri, Microsoft’s Cortana and Amazon’s Alexa — who are all primarily female. In an interview with NPR, Alex Spinelli, the head of the team who created Alexa, noted why he chose a female voice for the assistant.
“The idea was creating the Star Trek computer,” Spinelli said. As for why Star Trek’s male creator Gene Roddenberry chose to make the voice female, we may never know, though some researchers believe the underlying motive to be how we see women in society.
“Service work, domestic labor, healthcare, office assistants — these are all industries which are heavily feminized — and also often low paid and low status, with precarious work conditions,” Miriam Sweeney, a researcher with University of Alabama, told ABC News.
Being aware of this and other biases has done little to change our preferences. Despite knowing the subtle sexism of a female virtual assistant, for example, our ingrained biases still color our decisions: Both men and women simply trust a female voice-assistant more.