Yup, you read that right. A study released last month by the National Bureau of Economic Research suggests that algorithms make better hiring decisions than humans do.
The study, which was conducted by Mitchell Hoffman, Lisa B. Kahn and Danielle Li, observed over 300,000 hires across 15 companies that employ low-skill workers such as call center operators or data entry employees. The study required the companies to implement hiring assessments created by PeopleMatter that asked candidates a variety of questions about their technical skills, personality, cognitive skills, and fit for the job. In some cases, hiring managers were removed from the process, and hiring decisions were made by an algorithm that based the decision on the test results. In other cases, hiring managers used their discretion to override the algorithm’s suggestion for the position.
Before the study, the average worker hired across these 15 companies lasted just 99 days in the position. Retention rates are among the most important metrics to consider in recruiting, due to the high cost of hiring new workers—one study estimates that replacing a single employee costs companies an average of six to nine months’ salary, between recruitment efforts, training costs and productivity losses. Because of the cost associated with low retention rates, you’ll understand that it’s a big deal that this study concluded that retention rates increased by 15 percent when an algorithm chose the hire.
The study also showed that the use of human discretion—that is, the use of the hiring manager’s own judgement to overrule the algorithm’s recommendations—was strongly correlated with worse hires. The tendency to trust one’s gut feeling over a machine’s recommendation, also known as “algorithm aversion”, is a widely observed practice, even though it’s likely hurting business practices. “It’s human nature to think that some of that information you’re learning in an interview is valuable,” said Danielle Li, one of the researchers that worked on the study. “Is it more valuable than the information in the test? In a lot of cases, the answer is no.”
The study also determined that hires made by human hiring manages were not any more productive than the hires made by the algorithm, quelling the possible argument that human hires were of higher quality than algorithm hires, even though they stayed for less time in the position.
Besides increasing retention rates, using algorithms in hiring can also eliminate bias from the process and increase workplace diversity. “From a human perspective, we like people who are like us,” said Julie Moreland, senior VP of strategy and people science at PeopleMatters, the company that built the assessments for this study. “They’re not thinking about the job, they’re thinking ‘I can work with this person, I relate to them.’
Hiring a candidate based on how well he or she would mesh with the work environment and other employees is known as hiring for “cultural fit”, and it can heavily influence the hiring manager’s decision. “Similarity between the interviewer and interviewee—they’re from the same region, went to the same school, wore the same shirt, ordered the same tea—is hugely influential, even though it’s not predictive of how they perform down the road,” said Cade Massey, professor and researcher of behavior and judgment at the Wharton School of the University of Pennsylvania.
The problem with using cultural fit as a hiring tactic? Besides having no correlation to workplace performance, it can also be discriminatory and result in a lack of workplace diversity. “Every company vets its own way, by schools or companies on résumés,” said Sheeroy Desai, co-founder and CEO of Gild, a recruiting software company. “It can be predictive, but the problem is it is biased. They’re dismissing tons and tons of qualified people.” And because hiring for cultural fit often happens subconsciously, it’s probable that discrimination is imperceptibly occurring across many organizations. “Anybody that says they do not have bias in their interview is not being real,” said Moreland.
One question remains: if algorithms hire better than humans can, should humans be altogether eliminated from the recruiting equation? According to Mitchell Hoffman, one of the researchers that worked on the National Bureau of Economic Research study, the answer is maybe. “[The study] definitely suggests that more decision making powers should be given to the machine relative to the humans,” Hoffman said. However, Moreland advises against removing humans completely from the recruiting process. Moreland suggests hiring managers continue to conduct in-person interviews in order to pick up on human subtleties that machines cannot, but that they must also heavily weigh results from assessments when making the final hiring decision.
Much more research still needs to be done on algorithms before companies can consider relying on artificial intelligence for hiring decisions. For instance, the study conducted by the National Bureau of Economic Research only researched an algorithm’s ability to hire low-skill workers; the algorithm wasn’t tested on higher skilled jobs.
Even if recruiting efforts do become more and more machine-based, there will always be a need for humans to monitor data and the algorithms so we don’t become exceedingly confident in the data or overly reliant on the machines. “I think we need a lot more study of this, and there is more of a growing interest in the technical side of things and how to do this,” said Suresh Venkatasubramanian, an associate professor at the School of Computing at the University of Utah in Salt Lake City who researches the fairness of machine learning. “There are lots of things we can do.”
All in all, the National Bureau of Economic Research’s report is just the latest study to confirm what other researchers have been saying for a while now: robots, artificial intelligence and machine learning are becoming more useful in the workplace, and their prevalence will only increase in the future. Luckily, we know that humans and robots can happily coexist in the workplace!