Can Machine Learning Ever Remove Human Bias?

Machine learning finds candidates that would often be missed by recruiters. But technology still needs people to make it work.

Technology has revolutionized our lives in recent years. At the same time, businesses’ search for highly skilled, flexible talent has remained a constant challenge. Now, digital innovation is being applied to recruitment technology and talent management, producing ever more ways to screen large volumes of applicants and assess individuals’ likelihood of success.

Algorithmic assessments and machine learning are increasingly useful tools for recruiters. But might machine-based learning overcome the human error and bias that exists in the recruitment process? And how can this innovation be adopted without removing the magic human touch?

What Makes Algorithms Useful

Human oversight remains an invaluable element of selection and recruitment, but HR and hiring managers have always struggled with three key questions:

  • How can I link the positive traits in a candidate’s application to specific business outcomes?
  • Which of these outcomes should I focus on?
  • Can I make these predictions in an unbiased way that doesn’t harm the recruiting process?

Tech advocates say the answer is machine learning and predictive analytics, which removes human judgment. But some experts question how useful this approach to recruiting really is. After all, as was pointed out in a recent Fortune article, humans create these models, so the algorithms themselves are inherently biased.

But what creates a poor outcome from an assessment is not the algorithm itself, but the methodology behind the algorithm creation. If used properly, algorithms should remove human-introduced opinion. Instead of mimicking and reinforcing human bias hiring, the ideal algorithm will objectively predict a business outcome, post-hire. In this way, technology should improve the quality of the labor force.

Let’s take an example. A recent model created for a call center representative role revealed that candidates with call center experience were actually likely to perform poorly. This is counter-intuitive, but technology quickly spotted this, whereas the human mind would be unlikely to do so. In truth, what gets a person hired is not always what makes them good at their job. But algorithms, unlike people, are good at identifying the difference between the two.

Testing the Tests

So, machine learning undoubtedly has the potential to make huge changes to hiring practices. But using technology to assess the best fit for a role from basic application information and resumés is easier than predicting an outcome post-hire. In complex scenarios, computers can make mistakes, too. For example, if a company favored computer science degrees as highly predictive of employee success, an algorithm might turn away a disproportionate number of female applicants, given their relative scarcity in the field. That is where being able to ‘back test’ an algorithm is critical. An algorithm back test uses candidates’ data to score historical applicants and track their demographic distribution along with their subsequent success.

If HR is to set the agenda on people management, it must either staff up to handle [data] analyses itself or partner with people who can do the work.

Peter Cappelli
But never forget the human beings in this equation. As the Wharton School’s Peter Cappelli put it in the Harvard Business Review: “If HR is to set the agenda on people management, it must either staff up to handle [data] analyses itself or partner with people who can do the work.” When it comes down to it, however useful machine learning proves, you can’t get away from people. If organizations don’t fully understand the methodology used in creating algorithms, they’ll need to hire the experts that do.


Author: SHL UK