How AI reduces unconscious bias in the hiring process

Early AI tended to replicate unconscious biases. Today’s AI, however, can help reduce that bias.

How AI reduces unconscious bias in the hiring process

Unconscious bias in the hiring process has been a hot topic for some time. Back in 2003, for example, researchers Marianne Bertrand and Sendhil Mullainathan found that when candidates had identical resumes, those with very white sounding names received 50 percent more callbacks for interviews than applicants with very Black sounding names. These results, as well as the results of similar studies, have led to more focus on human bias.

Early attempts to use AI in hiring focused on removing unconscious human bias by removing humans from key parts of the screening process. Yet in many cases, researchers discovered their own unconscious biases repeated back to them by the algorithm. Often, these biases were baked into the data sets, giving early AI no choice but to repeat what it had been taught.

AI-enabled hiring software has improved as our understanding of bias has changed. Today, cutting-edge AI can more effectively combat unconscious bias with broader data sets and a focus on skills.

Broader horizons through larger data sets

When AI was first applied to the candidate screening process, it was often fed data from within a single organization. On the surface, this approach seemed sound: To determine what helped people succeed within a particular company, the algorithm would need to know how everyone in the company had fared.

The results, however, reflected conscious and unconscious bias as much as they reflected useful hiring information. Some companies, for example, found that their early AI recommended hiring from a handful of colleges or universities. On closer inspection, hiring teams realized that the AI was reporting a pattern it saw in the data. It was one that reflected the hiring teams’ own preferences for their respective alma maters rather than an actual source of skills and competences.

“The bias usually comes from the data. If you don’t have a representative data set, or any number of characteristics that you decide on, then of course you’re not going to be properly finding and evaluating applicants,” says Jelena Kovačević, dean at NYU Tandon School of Engineering.

For better insights on the connection between skills and job performance, AI needs a larger data set than that offered by any one company — even a company as large as Amazon, notes Jeffrey Dastin, technology correspondent at Reuters. By looking across massive data sets from a wide range of sources, AI can more easily differentiate the noise of bias from the signal of relevant skills.

Couple of young hikers are relaxing on big rock against sea and sky

Skills-based insights and predictions

Yet supplying larger data sets is just one step in improving AI’s ability to combat rather than replicate unconscious biases. To improve the hiring process, artificial intelligence tools need to focus on what matters to success in any given position, as well. They need to focus on skills and potential.

In the past, hiring teams had no AI algorithms or lightning-fast computers to help them analyze applications. All screening and analysis had to be done by human members of the hiring team, and it took time.

Human resources teams began to use factors like education or a sense of kinship with certain applicants as a shortcut. These shortcuts are not always connected to the skills necessary to succeed in any given role, much less to build a long-term career within an organization. They also tend to be sites of unconscious bias, which can be so deeply embedded within them that “simply undergoing ‘awareness training’ isn’t enough to eliminate them,” says Alexander Young, founder and CEO at learning solution Virti.

AI trained to focus on the connection between skills and career paths, on the other hand, can more easily focus on what truly connects an applicant to potential or actual success. Skills-focused AI can also be used to make predictions about where an applicant with a particular skill set might flourish, or what roles they are most likely to be able to grow into.

Early attempts to harness AI for improved hiring revealed much about how humans make hiring decisions. The more we learn, the more effective today’s AI has become at drawing connections between certain roles and the skills required to succeed in them. As a result, new generations of AI for hiring can help fight unconscious bias and spot candidates who might otherwise be overlooked.

Images by: antonioguillem/©123RF.com, cppzoe/©123RF.com

You might also like...