Biased AI Is Real; What Does that Mean for Women?

There are certain trends that will affect every person on the planet. Some more than others.

Climate change, of course, is a frightening trend that more and more people are becoming familiar with, as coverage of intense weather events and even disasters continues to scroll across news tickers and fill up our social media timelines.

Some trends are no less significant but have yet to permeate the public consciousness. One of the most important is the growing impact of the artificial intelligence revolution. AI’s effects on the daily lives of people around the world have the potential to be just as powerful as those of climate change. But just like with climate change, the mechanisms powering AI can be confusing to the general public.

This confusion can lead to a lot of bad outcomes, like people not taking problems in AI seriously, or denying that problems even exist. The fact is that AI experts are sounding the alarm about one important aspect of their field: biased algorithms. Recently, freshman congresswoman Alexandria Ocasio-Cortez pointed out that these algorithms can perpetuate society’s systemic inequalities.

“Algorithms are still made by human beings, and those algorithms are still pegged to basic human assumptions,” said Ocasio-Cortez. “They’re just automated assumptions. And if you don’t fix the bias, then you are just automating the bias.”

While some politicos and partisans jumped on these comments, the truth is that, just like climate experts agree there is a global warming trend and that human beings are the main cause of it, AI experts agree that biased algorithms are a major problem for the AI revolution. Consider these two facts:

The existence and widespread influence of biased AI has major implications for everyone living in the time of the AI revolution, good or bad depending upon whether the bias is in their favor. Facing systemic inequalities across the globe, women are more likely than men to have the cards stacked against them, and the AI field is no exception.

To see how this manifests in the real world, look no further than the workplace, where women have traditionally been discriminated against throughout history. Rather than fight this inequality, biased algorithms are actually helping to keep it going. One powerful example: an Amazon recruiting system that looked for patterns in resumes in order to surface the best ones taught itself that male candidates were preferable, actually downgrading resumes that included the words “women” or “women’s” or mentioned attending women’s colleges. Another example shows us how women are prevented from learning about good jobs in the first place, as a Carnegie Mellon University study found that Google ads showed high-income jobs to men at a much higher rate than it did to women.

Companies that produce AI products and services are waking up to the dangers of biased algorithms, with both Amazon and Google taking steps to rectify the problems above. In fact, a recent SEC filing from Microsoft warned investors about biased algorithms, stating that they could “undermine the decisions, predictions, or analysis AI applications produce, subjecting us to competitive harm, legal liability, and brand or reputational harm.” But biased AI is bigger than any one company’s bottom line. Without actively fighting these problems, artificial intelligence will become yet another structural behemoth founded in inequality that perpetuates bias against certain groups on a massive scale.

This month is Women’s History Month, a time to both celebrate how far we’ve come and push for even more progress. We have a great opportunity to lay a better foundation for the future of AI if we encourage more women and people of diverse backgrounds to dive into data science and other AI fields. And those of us already on the front lines should use this time to think about how we’ll fight against bias in our own domains. The AI revolution is set to change the world; let’s make sure it’s for the better.

Sophie Searcy is a Senior Data Scientist at Metis, which accelerates data science learning for individuals, companies, and institutions through corporate training and accredited, immersive bootcamps.