Many of us are at least a little scared of a future where algorithms dictate our decisions. And this fear is justified — algorithms present the grave danger of perpetuating existing human biases and inequalities while parading around as fair.
As we all know, humans have always been biased. Study after study shows that women and minorities are at a 1.5 to 2x disadvantage in hiring, promotion and other workplace processes. Recent Facebook, Google and Uber lawsuits belie the consequences of this reality. Technology, on the other hand, is neither good or bad; it only knows what we teach it. So we shouldn’t be surprised that technology — created and trained by inherently biased humans — is perpetuating our biases.
The good news is that AI is actually much more powerful than technologies we’ve created in the past. With humans, it’s virtually impossible to truly remove our inherent biases. We have attempted to improve this through unconscious bias training, and while this effort is well-intentioned, it has been shown to be patently ineffective.
While removing bias from a human is close to impossible, removing it from an algorithm is not only possible but already a reality. We can engineer and train algorithms to be bias-free. We can ensure that they treat individuals from every background equally.
How is this possible? Because the output of algorithms can be trained to be replicable and pre-tested to be bias-free. Humans, on the other hand, are both unpredictable and impossible to pre-test.
The potential for AI to promote equality is profound. Yet all we hear about is AI creating evil robot overlords that instigate dystopian futures full of inequity. Why are we so focused on portraying AI as ubiquitously negative? Look around — humanity is doing a fantastic job of perpetuating inequality without robot assistance. Actually, we could really use some AI help!
AI can be an incredibly powerful force for good. Instead of viewing AI as a dystopian overlord, think of it as a bias-reversal mechanism that can undo the powerful biases introduced by even the most well-intentioned humans. This makes it one of our most powerful tools for promoting diversity, equality, and inclusion.
Companies are starting to get with the program. Faced with the reality of lawsuits or heavy-handed regulatory initiatives like eliminating background checks in an attempt to promote diversity, they are more incentivized than ever to turn to bias-free AI. pymetrics, Textio, Applied, Unitive are all examples of platforms using technology to reduce bias. Each of these platforms has tangibly improved diversity outcomes for companies, helping them achieve record levels of diversity across gender, ethnicity and socioeconomic status. For example, at pymetrics, we’ve helped Accenture improve their gender balance in tech hiring by 20% and a major financial services company move their gender balance from 25% to 50%. And this is not at the expense of talent prediction; we also improved their quality of hire and retention rates by up to 60%.
Blind auditions for orchestras moved the needle in female hiring from 5% to 25%. We currently stand at 5% for many female roles – women CEOs, both Fortune 500 and startups, CTOs, engineers and a whole host of other male-dominated fields. What we need is the “blind audition” for the workplace.
AI can be scary, and when left unchecked it can definitely perpetuate existing disparities. However, if we are intentional about how we design it, AI can be our ultimate tool to create the “blind audition” we have been looking for.