A new generation of recruiting management systems promises to reduce human bias. These systems use machine learning and big data to discover what they claim are the best candidates. They use data, not intuition or first impressions, to recommend candidates for interviews.
This takes applicant tracking systems well beyond resume keyword scanning. To make their assessments, these systems use historical data, including past hiring decisions, as well as data on the candidates, to determine what constitutes a successful employee. The goal is to find patterns in this data.
Results based on data
A hiring manager, for instance, might believe that a job requires a particular skill, but recruiting management software may discover that this hasn't been true for the firm's top-performing employees. This result may be counterintuitive, but it's based on data. It means the qualifications the employer thought were needed may include some false assumptions. This may be a particular problem for the tech industry.
"A lot of times, these tech companies have traditionally focused on a very narrow candidate pool," said Ji-A Min, chief data scientist of Toronto-based Ideal, which makes what it terms a "recruiting automation system."
One thing tech firms sometimes do to narrow their candidate pools, Min said, is to favor graduates of certain universities or people who work at a competitor. Ideal's system works to widen these candidate pools and looks at those who match the qualifications.
Bias is an issue generally in candidate selection, but especially in the tech sector. Last year, the White House, under President Barack Obama's administration, concluded that diversity in science, technology, engineering and math has "been undermined, at least in part, by systemic barriers," and "prominent among these are both implicit and explicit biases," it said in a report. Implicit bias is also referred to as unconscious bias.
This data that recruiting management systems use can come from an analysis of resumes in an employer's databases, which could number in the millions, along with hundreds of thousands of hiring decisions. These large data sets are a perfect application for machine learning, Min said. "You can train the algorithm to learn what a good candidate looks like," she said.
"Our system doesn't tell you who to hire; our system tells you who you should be interviewing," Min said.
Early days for adoption
Recruiting management systems also can be used to help increase minority and gender representation. A company will stipulate that all candidates "have to meet the qualifications," but if they are a woman or minority, you can bump them up in the process, Min said.
It's still early days for adoption of this technology, but a recent report on AI by McKinsey Global Institute backs up its potential in recruiting. Machine learning-enabled recruiting management systems may be able to "pinpoint the precise skill sets and personal traits that would enable someone to be successful in a job."
"Artificial intelligence may also help detect promising candidates with less conventional credentials and free recruiters from using school reputation as a proxy to assess candidates' potential," McKinsey wrote in its recent report.
David Lewis, CEO of OperationsInc, an HR consulting firm, sees potential in machine learning-enabled recruiting management systems and believes it is "theoretically correct" that these machine learning systems can eliminate bias. But he also has some concerns.
A note of caution
"The technology is there to eliminate you -- not to try to figure out how you can fit," Lewis said. "And that's where I think this potentially goes off the rails."
Lewis' point is that, if recruiting management systems aren't properly adjusted for market conditions, it might eliminate people at a time when hiring is difficult. He pointed to the difference in demand during the start of the recession in 2009 and today when unemployment is low.
Lewis also said that many hiring managers don't have the training they need and act on intuition. That problem extends to technology use.
"Companies also don't understand how to either select the right technology or use the available tools that the technology offers to get the technology to do what they want them to do," Lewis said.
The vendors say these systems have to be routinely adjusted to account for the data issues and changing conditions. For instance, if a firm's employees are mostly male, that could impact the outcomes, said Xavier Parkhouse-Parker, co-founder and director of recruiting technology provider Plato Intelligence, a London-based firm. The way around it is to carefully select data that includes all socioeconomic backgrounds and ages, he said.
These systems don't eliminate bias in the hiring process, and a manager can reject a candidate for any reason. But with the data, "you get something that is really hard to argue against," Parkhouse-Parker said.