On some level, the vast majority of companies use it – but many are finding it produces some unexpected results. This is algorithmic hiring, and it’s one of the hottest topics in recruiting and HR right now.
Applications of Algorithmic Hiring
For the most part, algorithmic hiring is not happening as a standalone process. Instead, many organizations that are incorporating AI and algorithmic learning into their hiring are doing it as part of a larger process. The most common is the use of an algorithm to scan resumes and synthesize the data into a “score” of suitability for the role, which then is used to rank and narrow down the field of candidates who will be handled by a human hiring manager.
Critically, these decisions as to what keywords and other data to “include” are made through machine learning, not human programming. The AI “learns” from past data, such as the profiles and resumes of past and current employees, their job performance data, and so on. The algorithm uses this information to piece together a picture of what resume keywords correlate with positive job outcomes, then scores applicants based on that data.
Resume scanning is by far the most popular and commonly-used application of algorithmic hiring, but it’s not the only one. Some companies provide clients with proprietary assessments, such as quizzes and games that applicants engage in, and then provide scores based on how their AI determines that the outcomes of those exercises correspond with job-based traits.
Algorithmic hiring technology has been adopted to varying degrees and at companies both large and small. 99% of Fortune 500 companies use some form of “sifting” software in the hiring process, including giants like Facebook, Google, Deloitte, and Nestle. Smaller companies, too, are using these solutions, though they may lack the resources to monitor and vet the results as closely as larger teams.
Advantages and Challenges
Proponents of algorithmic hiring tend to focus on its ability to save human hours, as well as the possibility of using it to work on reducing bias in hiring. Both of those advantages, however, come with corresponding challenges.
By having an algorithm handle the initial intake of applications, companies can save the time it takes for a human to manually sort through what may be a vast number of applications – many of which may be enough of a mismatch to immediately discard. Instead, those human employees can get those hours back to do the tasks that truly need a human touch.
That being said, algorithmic hiring is not a cure-all, and it can introduce new problems even as it solves others. Despite the popularity, 88% of employers believe that their automated systems actually remove qualified applicants from the process because they don’t meet some specific criteria that the algorithm values.
The hope is also that algorithms can also help remove some of the biases in hiring today. Humans are prone to a variety of biases and even our unconscious biases can affect our decision making. Despite consistent public pushes for improvements in diversity, equity, and inclusion (DEI), many organizations have failed to address the root causes limiting the acquisition of diverse candidates. Some hope that the use of algorithmic hiring can help to remove unconscious bias from hiring decisions and, hopefully, increase diversity in hiring. On the other hand, there is a strong possibility that algorithms, instead of eliminating human biases, may simply reproduce them. Algorithms simply “learn from” and attempt to reproduce past (human) decisions, which means that bias in that dataset are likely to reappear in the algorithm’s decisions, starting from job advertisements and continuing through the entire process.
For instance, a 2018 report revealed the disastrous results of an Amazon experiment with algorithmic hiring and screening. Although the algorithm was supposed to be rating candidates for software developer and other tech jobs, it eventually “learned from” an existing pattern of male bias in the tech industry, teaching itself to “downgrade” resumes that mentioned women’s clubs or women’s colleges, to prioritize resumes that included descriptors more commonly found on men’s resumes, and to assign minimal value to actual skills like various coding languages. Another company, in the process of vetting an AI tool for potential use in hiring, found that the top two factors the algorithm “learned” were indicative of job performance were totally irrelevant (and fraught with subtle bias): playing lacrosse in high school and being named Jared.
Although the human team overseeing the experiment kept trying to remove those particular keywords from the algorithm, it could not guarantee that the AI would not keep finding new ways to replicate old biases. Something as simple as the name of a university or the mention of a hobby or club could trigger these biases, and it’s impossible to keep up with them all. Playing whack-a-mole with those issues and keywords would certainly not be an effective use of anyone’s time or budget.
Why Partner with a Recruiter?
Algorithmic hiring seems to work best when it is only one part of a process or used as an “assist” to human-based recruiting and hiring processes. In fact, some of the benefits of using AI can also be achieved by partnering with a recruiter. For instance, if one of the primary concerns is stretching a team too thin, a recruiting partner can take that work off their plates and ensure they move forward with the best-fit candidates. Recruiters can also source candidates from a pre-existing pipeline of vetted, qualified individuals, significantly reducing the need for algorithmic evaluations in the first place.
Recruiters can also help to uncover “hidden workers”: those highly qualified candidates who are nonetheless eliminated by a quirk of the AI. Companies that intentionally seek out these workers are 36% less likely to face skills and talent shortages, compared to those that don’t, and an expert recruiter can help locate the best of this talent pool.
At any level, an expert recruiting partner can help to navigate the choices made with the algorithms, hopefully adding a “human touch” back into the process and evaluating the AI decisions in light of concerns over DEI and other factors. Recruiters have long-term experience and wide-ranging connections, and those human connections are one aspect that can never be replaced by an algorithm.
By Ruben Moreno
About the Author
After a 25-year career in Corporate Human Resources and HR Executive Search, Ruben Moreno and his two partners co-founded Blue Rock Search based on a simple but ambitious vision of creating a firm that would “Change Lives and Organizations One Relationship at a Time.” Ruben leads the Blue Rock HR and Diversity Executive Search practice specializing in the identification, assessment, recruitment, and onboarding of Chief HR Officers and Chief Diversity Officers and their respective teams — inclusive of leaders in Talent Acquisition, Total Rewards, HRBP’s, Learning & OD, HR Technology, HR Operations, and HR Analytics. Ruben has helped place hundreds of HR Executives and built deep relationships within the CHRO community across multiple industry verticals. His clients consider him a trusted partner who takes the time to understand their business and add value beyond executive search.