Written by Chris Russell
28th April, 2021
What is Automated Resume Screening?
How much time does a recruiter spend reviewing a resume? Would you believe mere seconds? 7.4 seconds to be specific, at least according to a 2018 study by The Ladders.
With every recruiter handling a dozen or more openings and every opening getting 50 or 100 or more resumes, there’s little choice but to quickly scan.
At least, that’s the way it used to be. Now recruiters at companies all over the globe rely on screening tools to review, score and rank the incoming applications and resumes. These tools free them to look more closely at those candidates that come out on top. They also can eliminate the bias that might unconsciously influence a human recruiter.
The earliest of these screening tools merely compared resumes against the requirements of the job description. This keyword matching ranked candidates by how many of the requirements were present in their resume.
AI Resume Screening
Today, AI resume screening is the state of the art. Going far beyond the keyword matching of an applicant tracking system, these tools understand context, and can impute good communication skills to a candidate whose resume mentions conference presentations and written work. Many will supplement a resume with information gathered from social media and public profiles.
The most valuable – and controversial – feature of AI screeners is their ability to learn a company’s hiring preferences. They do this by analyzing existing company records for patterns among those who were hired. If a company tended to hire more people who had worked for a competitor, candidates with that background would score higher than those without that experience.
When performance information is included, AI systems can perform even better, looking for those candidates who most closely match the company’s best workers.
Because preferences change, AI screening tools do, too. Their machine learning algorithms constantly update the scoring system based on who now is getting interviewed and hired. The more data these programs have to work with, the better their scoring and ranking.
However, if the data used to train these systems about hiring preferences is biased, then the screening criteria and scoring will be too. Amazon’s company-built AI recruiting system may be the most famous example of this.
After bringing its program online in 2015, Amazon discovered the system was biased against women for technical jobs. Programmers had trained it using a decade’s worth of tech hiring information. Since most of the software developers were men, the system learned to prefer male candidates. It penalized resumes that included gender-suggestive backgrounds and phrases.
After trying unsuccessfully to eliminate this bias and finding unqualified candidates were being recommended due to other problems with the training data, Amazon shut down the project.
Builders of these screeners have learned from the Amazon experiment and now routinely audit their scoring and selection systems. The data used to train these screeners is also scrubbed of details like names, gender and racial references and even suggestive phrases.
Some vendors of these programs have turned to skills and personality assessments to predict candidate success. They work by comparing how candidates perform on these assessments to the performance of the company’s best workers.
Though this selection method is less likely to discriminate on the basis of sex or race, it can result in hiring workers who all think and act alike, limiting the potential for creativity and problem solving.
There’s no question that AI candidate screening is a powerful tool to help recruiters find excellent candidates they might otherwise overlook in a 7.4 second resume review. When audited regularly and managed by talent acquisition professionals sensitive to the issues, AI resume screening can lead to hiring better workers more quickly and more efficiently.