How machine learning can evolve hiring for the better
The main solution to closing gaps in hiring? Implementing less rigid recruitment practices.
Very few people considered the possibility of the post-pandemic world being a candidate-sided labor market, but we now find ourselves exactly there. Many have called this time “The Great Resignation,” and understandably so. Turnover is high, and it’s becoming increasingly difficult to find new candidates to fill a myriad of open positions.
There are now more jobs available than there are candidates, creating a strain on HR teams to find top talent quickly. In fact, Julie Labrie, president of BlueSky Personnel Solutions, has noted that finding the right candidate can take as long as three to four months.
Related: Recruiting challenges drive increase in stress for hiring managers
This year, Harvard released a report outlining the various reasons why companies are facing difficulties in the labor market, including where the gaps lie and how we can address them. The main solution to closing these gaps is by implementing less rigid recruitment practices.
In their research, Harvard discovered there are an increasing number of “hidden workers” who are filtered out unintentionally during recruitment. When companies are focused too closely on job-specific criteria, they lose out on candidates that would otherwise be the best fit for a role.
Automation and artificial intelligence for inclusivity
Traditionally, companies have screened candidates based on a rigid set of expectations, all of which fall under what Harvard refers to as “negative filters.” These filters include education, gaps in employment history, and experience, among others. But assessing a candidate based on these standards can actually foster a less inclusive recruitment process.
In terms of automation, ATS systems are designed to make hiring more efficient. It is only natural for HR teams to want to instill faster hiring practices to filter through candidates. However, these systems can be too rigid in terms of selecting specific candidates. This is because ATS systems are trying to match candidates within the limited barriers set out in a job description.
Additionally, rigid ATS systems foster deeper exclusivity for positions. A candidate may not meet the exact criteria of a job description due to a lack of previous opportunities, but they still might be the right candidate. In a world where we are constantly seeking to increase our diversity, equity, and inclusion (DEI) goals, we should be working to implement hiring practices that help foster these goals.
In terms of filtering out the best candidates, traditional ATS systems that focus on specific criteria don’t consider the skills or behaviors candidates possess or exhibit that would make them the best fit for a role. However, with new and innovative technologies, we can now apply machine learning to identify these traits, surfacing the top talent while meeting the initial recruiting goal of efficiency.
By creating candidate behavioral assessments using I/O Psychology and machine learning, many companies are shifting their recruiting methods to focus on skillsets that make a candidate great. Harvard says that AI can be used to identify what makes current employees successful and apply their findings to “a new and powerful framework—hiring on the basis of skills and demonstrated competencies, not credentials.”
Behavioral assessments do just this. By focusing on a candidate’s skillsets, we can switch from hiring based on the “negative filters,” and focus more towards “affirmative filters,” as Harvard describes.
Machine learning is a powerful tool, and when used correctly, we can not only find the right talent quickly, but also without bias. By using unbiased algorithms to teach AI, we can look at identifying specific behaviors and skills that make a candidate right for the job. In doing so, we can also find the right candidate without ever considering gender, ethnicity, race, sexuality, or other unconscious biases—something all humans possess—that influence our decisions.
Read more: