The ‘Discriminating Systems’ report explores how inherent bias in AI could cause major future problems
New research reveals that the artificial intelligence field is dominated by straight, white men. According to a monumental new study, this singular subset of society is teaching intelligent machines to create the algorithms that will define our future, and their perspectives are becoming dictums that could threaten people of colour, women, and the LGBTQ community.
The Discriminating Systems report was published yesterday by researchers at the AI Now Institute at New York University. Their findings reveal that women make up just 15 per cent of AI research employees at Facebook, and only 10 per cent of the AI workforce at Google. In an even starker divide, the total of black workers at tech companies including Google, Facebook and Microsoft range between 2.5 and 4 per cent.
“This is not the diversity of people that are being affected by these systems,” AI Now Institute co-director Meredith Whittaker told press. “There is an intersection between discriminatory workforces and discriminatory technology.”
Researchers believe the clear bias towards employing white, cis men in tech demonstrated by these figures is provoking “a diversity crisis” in artificial intelligence. The report is the result of analysing over 150 previous studies, and cites cases where AI algorithms mistook black people for gorillas, in addition to the trans employees who faced discrimination at Uber after recognition software failed to identify them. The research ultimately found that “bias in AI systems reflects historical patterns of discrimination”.
Artificial intelligence already shapes our lives - these algorithms can determine who gets a loan approved, who gets that job interview, and who the police choose to keep tabs on. The way they’re taught to identify and categorise us will have a profound impact on the future of society.
Attempts to even the balance in the tech workforce by adjusting the pipeline of potential employees have made little difference. “Existing methods have failed to contend with uneven distribution of power,” said Kate Crawford, co-founder and co-director of AI Now. “Fixing the so-called pipeline problem is not going to fix AI’s diversity problem. We need to know that these systems are safe as well as fair.”
Researchers believe that the focus on widening the pipeline of available employees ignores more fundamental issues within the companies themselves. Tokenism, biased hiring models, and workplace harassment all contribute to the marginalisation of underrepresented groups in tech.
Instead, researchers say tech companies must commit to more practical responses, and make data on recruitment processes, discrimination, and wage distribution more transparent. The report also highlighted the need for more clarity in how AI systems are developed, calling for more thorough tests to include pre-release trials, independent auditing and continued monitoring.
“The diversity crisis in AI is well-documented and wide-reaching,” the report concludes. “It can be seen in unequal workplaces throughout industry and in academia, in the disparities in hiring and promotion, in the AI technologies that reflect and amplify biased stereotypes, and in the resurfacing of biological determinism in automated systems.”
Ultimately, AI is a malleable technology. If humans mould it to reflect outdated prejudices, those views will continue to shape society. The tech workforce must become more inclusive to reflect the diversity of the people its algorithms influence.
You can read activist Janus Rose’s view on how AI recognition software could threaten trans identity and why giving AI assistants human voices is an ethical nightmare on Dazed.