What AI Can Teach Us About Diversity
In 2015, I came across an articleabout a team at Rutgers University using an Artificial Intelligence algorithm to judge creativity in art. In 2016, and AI algorithm called Beauty.AI judged an online beauty contest and nearly all the winners were white. Last week, I read an article about researchers at Auckland University who have developed an AI algorithm to improve the response to child abuse reports. It’s obvious that if the algorithm is biased, the outcome of each of these projects will be biased. It’s no surprise that AI-driven decisions are reflect the experiences and world views of their creators.
I’ve been thinking a lot lately about biased algorithms. When I scan the shelves looking for books about leadership, the majority of volumes are written by men. For the past three years, there have never been more than four (out of ten) books written by women listed on the monthly New York Times bestseller list in the business category. The average is 1.5. Even fewer are written by people of color. Is it any surprise that our collective “cultural algorithm” of a successful leader looks like a white male? In 2017, Beauty.AI introduced Diversity.AI to address the baked-in bias in the first contest. Others are looking to increased transparency to address public accountability for AI decisions. Nonprofits and small businesses may not (yet) be using AI to drive their work, but we should not ignore the lessons we are being offered about baked-in bias and diversity.
All organizations have systems and policies that have the potential to perpetuate widespread and long-term bias. Most of my nonprofit clients want to create leadership pipelines in their organizations. Some have implemented mentorship programs or formal management training opportunities to encourage staff to take on new challenges and to create a talent pool for promotion. These initiatives can be valuable opportunities to help staff develop new skills and grow in their roles, but they can also perpetuate the organizations’ baked in biases. I have seen organizations exclude part-time employees or those who are new to the organization from participating in these programs. While there are good reasons for these exclusions, they can also result in missed opportunities for the employees and the organizations. Systematically excluding part-time employees can systematically exclude women who have returned to work after maternity leave or employees who are enrolled in college or postgraduate programs. Excluding new employees means that some people will leave the organization before being given a chance to participate in leadership development programs. Both of these exclusions very often leave out women, younger workers and people of color. Organizations leave a lot of talent on the table when they implement narrow algorithms that reflect institutional bias. Investing in a thoughtful a process to assess baked-in bias can lead to lasting value for any organization and can have a proportionately big impact on smaller businesses that need to maximize all the talents and resources at their disposal.