WHY THIS MATTERS IN BRIEF
The same technologies that are helping us catch and profile criminals, or surveil civilians at scale, AI, biometrics, machine vision and so on, are also helping companies find, interview, analyse and categorise job seekers faster and more effectively than ever.
Companies, like RBS and the Chinese state railways, are already using rudimentary Brain Machine Interfaces (BMI) and mind reading technology, yes, really, to assess potential candidates and monitor their workforce, so it’s no surprise that Artificial Intelligence (AI) and other related technologies are starting to make their way into the recruitment process.
“It felt weird, like I was kind of talking into the void,” said Sarah, a 27 year old marketing manager from Ohio when she first used HireVue, an Artificial Intelligence (AI) powered on demand video interview platform for job seekers.
The human recruiter she was working with told her it was “just like an interview on Skype,” so she followed the interview tips on the company’s website, making sure she was dressed appropriately and had a well lit background, but to her surprise when the “Skype like” interview started she wasn’t face to face with a person, she was face to face with an AI avatar. No humans in sight, and funnily enough her recruiter had never mentioned that the interview would be being held and analysed by an advanced machine learning platform, and that everything from her facial expressions to her word choices would be evaluated by a series of algorithms.
“You usually have a little time to do some small talk, but in the HireVue interview, I only had a practice question and then just went into it. There’s not a lot of time to feel ready,” she said of the interview that took place early last fall, “for me first impressions are everything, and it was hard to set that tone.”
It must have worked, however, because she got the job offer.
Of course, using AI and biometric feedback to monitor and analyse people and their behaviours, responses and reactions to things is nothing new. After all, privacy busting surveillance systems, which have come on leaps and bounds in the past few years, do that day in and day out, but as AI’s machine vision capabilities continue to improve, being able to do everything from judge a person’s character and their predisposition to criminality, to being able to analyse writing behaviours and styles, and detect when someone’s lying, as well as the smallest of micro facial movements and the faintest inflections in a person’s voice that give away tells, including if they have an undiagnosed illness or disease, any one of which could then be used as an excuse to pass you over for a job, it’s fair to say that this tech is going to soon be much more powerful than anyone realises.
As a result the potential power of this technology, which is now increasingly being used to autonomously interview and scan hundreds of thousands of hopeful candidates at scale, without human intervention, is mind boggling.
Throw in the much talked about algorithmic bias and the black box nature of these AI’s and all of a sudden it’s easy to see how quickly companies could find themselves, albeit behind the curtain of being able to increase “recruiting productivity,” companies could quickly find themselves in a tenuous position. Imagine, for example, an individual claiming that the process is biased against them, and suddenly you can see the class action law suits starting to stack up.
Today about 38 percent of working Americans are apparently actively looking for a new job, or plan to sometime this year, according to a recent report by Glassdoor. But, like Sarah, they might be surprised to find that those “first impressions” so carefully emphasised by career coaches are now being outsourced to an AI.
A 2017 Deloitte report found 33 percent of survey respondents already used some form of AI in their hiring process, and with jobless rates at a 17-year low of 4.1 percent according to a February report from the Bureau of Labor Statistics, recruiters are increasingly looking for ways to bring in the best candidates faster and more efficiently than before. And an emerging crop of new, “smart” hiring tools can do just that by cutting down interview processes from what traditionally took weeks to a matter of a few days.
Some of the new tools involve as little as answering a text message. In 2012, while building his company FirstJob, an online job board for millennials, Eyal Grayevsky discovered that many candidates never heard back from employers and that their materials seemed to go into a “black hole.”
Four years later Grayevsky launched an AI recruiting tool named Mya, short for “my assistant,” and rebranded his company as Mya Systems. In Mya’s case they help company’s recruiting process by directly engaging with candidates via text, asking basic questions such as start date and salary requirements. Candidates can also ask Mya questions, and when she doesn’t know the answer, she will query the recruiter.
Within minutes Mya rules out candidates based on a pre-programmed assessment model or moves them along to the next part of the interview process. The experience is so seamless that 73 percent of surveyed candidates who had interacted with Mya reported they had interacted with a recruiter when they in fact had spoken only with the bot.
“Now 100 percent of candidates are getting a response, everyone is getting a chance,” said Grayevsky, “candidates feel like they really get a chance to express themselves to the company with more than just a résumé,” where, of course “the company” is a deterministic autonomous bot.
AI powered video interview hiring tool companies, like HireVue and Montage, also boast speed as a key for why more and more recruiters are relying on their services. On demand video allows candidates to interview any time of day, and in turn, once they’re filtered by the AI, recruiters can compare dozens of interviews, all in the time it takes to commute to work.
“The most overused metaphor is that there’s a war for talent right now, but it’s actually not a war, it’s a race,” said HireVue CEO Kevin Parker, “and the people that are the fastest-selecting and reaching out to candidates are the people that win and enjoy a competitive advantage.” And he has a point.
As for candidates who feel more trepidation than empowerment from video interviews, HireVue offers tips on its website and frequently engages with users on Twitter about how to best handle the hiring process. Many of the suggested steps are ones that interviewers will be familiar with, such as researching the company, practicing and preparing for common interview questions and dressing appropriately for the job.
Where on-demand interviewing differ though is that you should also practice your facial expressions and exaggerate them, for example, a huge smile that might seem ridiculous in person will be picked up more easily by the AI. You’ll also need to make sure you have a good internet connection and bright lighting, and get some of these “new” interview techniques wrong and you could find yourself easily being rejected for the job you’re interviewing for.
One question they get frequently, says Lindsey Zuloaga, who’s the Director of Data Science at HireVue, is if an applicant can trick the AI.
“If you can game being excited about and interested in the job, yes, you could game that with a person as well,” she said, “you’re not going to game it without being a very good actor.”
Another big advantage for HireVue, apparently, is that it offers a customizable AI to help assess candidates’ video interviews. The AI gives each video interview a score based on more than a staggering 250,000 data points that include audio, tonality and speech patterns, and much more, and the importance of which can all be customised based on the client’s needs, and because it’s based on machine learning, HireVue’s AI can learn and refine its accuracy over time based on new data.
Jim Cochran, head of global recruiting at JPMorgan Chase, who also began automating their law teams, said that the process of working with HireVue to build an AI that matched their recruiters’ needs took about a year, with a substantial part of the process geared toward evaluating the factors that best target a successful employee within the general population.
But after some tweaking, he said their recruiters have been happy with the results so far and are planning on working with AI modules for more positions. Though JPMorgan has been a HireVue customer for four years, the company believes that adding AI has helped speed up the process of filtering through videos.
“It’s unstructured video and audio coming in, and this is a way of structuring [it],” said Zuloaga of HireVue, “it’s just kind of hard to get the information that you really want to know about a person from a résumé or a multiple-choice assessment.”
Zuloaga works with the company’s industrial psychologist to make sure the product’s assessment tools are up to industry standards, but adding AI to the mix gives the tool an important advantage – rooting out bias. Although, I’m sure there are others who’d disagree, as the “Bias in AI” debates rage on.
“We can measure it, unlike the human mind, where we can’t see what they’re thinking or if they’re systematically biased,” said Zuloaga, “if the team does notice a skew in results, we can evaluate the algorithm to see what went wrong and remove the bad data.”
And while there’s no guarantee that AI will completely eliminate bias in hiring, especially once the candidate reaches a human recruiter, companies using HireVue have reported a much more diverse candidate pool. Parker pointed to Unilever, which has improved the diversity of its talent pool by 16 percent since partnering with HireVue.
Grayevsky said Mya’s customers have seen similar results.
“It’s really easy [for recruiters] to go to the applicants that feel safe or the ones they recognise, whether it’s the school or the types of companies they’ve worked at in the past, but Mya really is only interested in who’s active, who’s interested,” said Grayevsky.
HireVue and Mya are just a few of a growing number of companies looking to make their mark on a recruitment industry that is valued at up to $200 billion and growing. TalVista, another California based start up, seeks to eliminate bias from job descriptions by using A.I. to tweak the language so that it’s more appealing to women and minority applicants.
In addition, to help automate the recruitment process, Entelo analyzes a candidate’s social media presence to determine their fit for a position.
Although AI can speed up the process of getting the right candidate in the door, in professional industries with limited job seekers, making candidates aware of positions in the first place presents a larger hurdle. In fields such as nursing, IT, and middle management, Mya serves customers in these categories by actively reaching out to candidates already in their application system, alerting them to new opportunities. Grayevsky says the company is also in the process of launching a partnership with several job board sites in order to further widen the pool. Ultimately, his goal is to “create a scenario where candidates know to reach out to Mya for support” before starting their jobs search.
HireVue is also interested in working with companies improve their internal matching for open positions. The company already offers analysis of predictive job performance, something that the machine learning can only refine through updated data.
Parker says that they also want to help companies directly match applicants, especially recent graduates, with positions based on assessments rather than relying on traditional job listings that might miss the right candidate.
Games or simulations to help candidates get a better sense of a job position are also gaining popularity, with 29 percent of global business leaders using some version of the technology according to Deloitte. Kurt Heikkinen, president and CEO of Montage, which works with a number of Fortune 500 companies, stressed the important of these kinds of highly branded, personalized experiences in an age of competitive hiring.
“Through technology candidates behave much more like consumers, they want and deserve convenience at their fingertips,” he said.
But convenience isn’t always enough. Cochran expressed concerned that, without proper follow-up, on-demand videos could turn into another “black hole” for candidates.
“I’m very focused on making sure this additional step we’re asking them to take is met with a very responsive recruiter or recruitment system,” he said.
And while Sarah felt a little thrown off by her first HireVue interview, she said she plans to go back to the same recruiter for her next job.
“Facial recognition is just everywhere. If I can just put on some makeup and that adds a couple points to my score, I’m not going to be mad about that,” she said, “I think it’s just making sure candidates are informed.”
But facial recognition is just the tip of the iceberg, and for those who are interested in the future trajectory and development of these platforms then my advise is look at what’s happening in the surveillance and privacy fields, and then watch as all those technology developments slowly, and sometimes silently, creep into these platforms. Companies, and candidates for that matter, should be careful that these systems don’t become biased, because as and, probably when, they do some candidates and demographic groups could be in trouble…