A.I. is playing a new role in hiring.
“Congratulations, you have been selected for an interview for the professional minigamer position at Open Mind Corporation,” a robotic voice announces over a blank screen. “My name is Alex. I will be guiding you through the interview. The whole process will take no more than 10 minutes. Let’s hear your voice. … Smile for the camera. … Welcome to the interview.”
This is the start of An Interview With Alex, a dystopian online interactive experience taking viewers through a “job interview” conducted by an A.I. hiring manager—one that measures tone to score users on a “State of Mind Index.” Carrie Sijia Wang, the multimedia artist behind the project, writes that her work is meant to “criticize the present by speculating about the future.” But it’s not that far off how your next job interview might look, if you’re applying for high-volume, low-skilled roles (or even some high-skilled ones). A growing number of real-life recruiters are turning to A.I.-led job interviews, using programs that interview and assess candidates before a human recruiter even lays eyes on them.
Alex won’t be interviewing you, but it might be Hubert, or Ella, or Tengai, or Phai, or simply automated words on a screen. Most of the bots aren’t running the decision-making process from end-to-end (although sometimes they are—see Ryan Fan’s OneZero article “I Got a Job at an Amazon Warehouse Without Talking to a Single Human”). Instead, recruiters generally use A.I. at the “top of the funnel” to sort or rank candidates before they reach a still human-run stage. Like humans, these bot recruiters have their own unique styles of interviewing. Some are merely seeking logistical information, like availability and ongoing interest, while others might be looking to assess drive, initiative, team-working skills, adaptability, or even your tendency to job hop. Some will ask everyone the same set of questions in the same order, while others will tailor their questions to you, verifying that you can really do the things you say you can.
For many positions, every single applicant receives an automated interview link. Applicants are free to log into the “interview” on their own time, and there will often be practice questions they can try before they face the official questions. Some are text-based, while others require applicants to record themselves on video. Questions, posed by a bot or prerecorded message (or, in one case, a disembodied head), are usually fairly stock standard probes: Tell us about your previous experience, why are you interested in this company? But there is no human listening. The answers are recorded and analyzed by A.I., marking the candidate’s suitability on certain traits, before human recruiters use this analysis to decide whom to invite to another interview or hire. By the time answers are reviewed by a human (if at all), A.I. has already passed judgment.
Odds are, if you’re applying for the kinds of jobs that attract lots of applicants, you’re going to be interviewed by one eventually. It’s so prevalent that Fast Company ran an article telling readers about “4 things you must do to prep for an AI-powered job interview,” while LinkedIn runs a free A.I. video interview practice tool. And as much as the companies behind them tell me otherwise, you might not love it.
Artificial intelligence has played a growing role in recruiting and hiring for some time, as both a timesaver and a matchmaker. A.I. has been used to generate job descriptions, to post and share jobs, to automate candidate searches, and to scan résumés and cover letters, leading some to hide keywords in white text on their applications. (Don’t do that.) Candidates may not have even noticed this shift, and perhaps it makes very little difference to most whether a job is promoted by a faceless HR manager or a faceless bot. But it’s one thing to have A.I. “read” your résumé—it’s quite another to have to answer its questions.
These systems are now being used by major companies, including Unilever, Vodafone, Intel, L’Oréal, Mars, and Citibank, to name but a few. Kevin Parker, CEO of Utah-based HireVue, one of the more prominent platforms in the space, tells me that one of the company’s customers, “a large grocery chain in the U.S.,” used the platform to interview about 20,000 people a day for stocker and cashier jobs. But A.I. interviews are also being used for internships and professional positions, especially in the age of social distancing.
The startups behind these platforms are all quick to point out their unique ability to improve efficiency and fairness, cutting out human bias and inconsistency. A.I. can avoid many human interviewer foibles: It is immune to charisma and distractions, never gets tired or cranky, and will remain on-script, ensuring consistent interviews and meritocratic decision-making. Plus, this way, 20,000 candidates a day can be interviewed and assessed by the same “interviewer.” While efficiency features prominently on these platforms’ websites (PredictiveHire has a free tool calculating how much time and money automating interviews can save), almost all of them make some mention of eliminating unconscious bias, overcoming societal preferences for certain races, genders, ages, and appearances. But algorithms have often been found to replicate human bias, especially if fed with human data sets, and certain characteristics may be unfairly associated with hirability. (In 2015, Amazon abandoned an “experimental” A.I. recruiting tool, which analyzed résumés rather than interviews, because it was discriminating against women for technical roles based on previously skewed hires. It preferred “masculine” words to feminine ones.) A.I. recruitment companies argue that with the right amount of auditing of data, these biases can be drawn out and removed. “We can look to see how the questions impact different groups,” says HireVue’s Parker. “Are men scoring differently than women, are Asians scoring different than Caucasians, are African Americans impacted differently? There’s a whole series of tests and statistical analyses that we go through to make sure that the algorithms aren’t introducing bias.”
Many of the companies behind these systems also market themselves on their “candidate experience.” While the hiring company is ultimately their true customer, A.I. firms are acutely aware of the impact a bad interview experience can have for a company, especially one that wants to keep its failed candidates as customers, and use this as a selling point. Hubert+1, a Swedish startup, notes that “a full 11% of candidates with bad experiences will sever all business relations with the company in question” and boasts that its “Hubert,” on the other hand, provides “a candidate experience fit for a king.” (As though kings are applying to jobs!) PredictiveHire, an Australian company that runs chat-based A.I. interviews, shares positive candidate reviews on its website. Parker says that HireVue’s average net promoter score, measuring if the candidate had a good experience, is “somewhere in the high 60s, low 70s”; Hubert+1’s website brags of 92 percent candidate satisfaction, PredictiveHire of 99 percent. “And if I gave you access to any hundred interviews and asked you to watch them with the sound off well, what you’d probably see most often is people smiling,” says Parker.
But I wanted to hear more from the people who have actually done one of these interviews—and there are quite a few complaining about the process online. To be fair, these anecdotes may be skewed: People who had perfectly fine experiences aren’t likely to tweet about them. It’s clear, however, that many find it difficult, disrespectful, and dehumanizing. Reddit is full of rants about HireVue, one of the most prominent platforms in the space, with posters calling it “the worst interviewing experience” or “a pure waste of time.” (Others use more colorful language.) Many say they won’t be taking them again. “This is a hard pass,” wrote one user.
Lauren, a 23-year-old college grad who has been applying for jobs since May 2019, told me that getting one of these interview requests isn’t the same feeling as landing a real interview. “You get the email and it’s like, ‘Oh my gosh I landed an interview!’ ” she says. “And then you realize it’s an A.I. thing and you’re like, ‘OK. Sigh. I guess we’ll just do this and hopefully get an actual interview.’ ”
Part of the problem is that bot interviews create more work for applicants, asking even those who are unlikely to get hired to invest the time. Jade, a recent college grad looking for work in retail, told me: “We have to do all this song and dance for A.I., or a robot, or whatever it is before we even talk to an actual person—if that ever happens.” Aurora, a job-hunting friend of mine who worked in the travel industry prior to COVID-19, says this addition takes up “so much time.” Despite being done alone and from the comfort of your own home, for video interviews, you’re still encouraged to look presentable every time—whether because the A.I. will judge now or a human reviewer will later. “You have to do your hair and your makeup and your clothes and get your background right,” she says. “That makes it extra exhausting, because you’re doing that every couple of days, just so that you can be seen on a screen in a professional manner.” (When I pointed this out to the A.I. firms, they were quick to argue the opposite: that this gives everyone a chance and allows applicants to complete on-demand interviews at their convenience.)
Others saw a different problem with A.I. sorting the initial pool: that it is being used to eliminate people before applications even reach a human. “It’s almost like they are trying to trip people up, that’s the feeling I got with a couple of them,” Aurora says. “That they were trying to get rid of people by making them have a certain kind of answer.” Lauren also got the feeling the A.I. was just waiting for her to slip up. “One wrong thing and the A.I. could just kick you out. … Or something. I don’t know how they work.”
One person who does understand how they work is Kat, a 33-year-old software engineering student who was recently offered two different HireVue interviews in her search for an internship. She tweeted about her concerns as a dark-skinned Black woman, because A.I. is known to perpetuate bias against people of color or fail to recognize them at all. She says most who responded to her viral tweet advised her to decline the interviews, although a number of recruiters showed up to reassure her. “One of my Twitter buddies, she actually does ethics and A.I., and she was like, ‘This is … this is … like, no. Absolutely not,’ ” she says. Kat did do the interviews in the end, against most of her friends’ advice. What put her mind at ease, she says, was learning that not all recruiters that use HireVue to automate questions actually use the A.I. tool analyze the answers. (Some merely use it to interview en masse, with human recruiters still watching and scoring them.) But she was disappointed to later read the terms and conditions and realize one of her interviews did use A.I. analysis; she plans to decline the next time she receives one. “It just felt like I was not valued as a human,” she says. “It just makes me feel like a collection of data points.”
But even when the answers aren’t analyzed by A.I., the automated interview process was still dispiriting to the candidates I spoke with. “It’s all one-way,” says Kat. “I think a lot of these recruiters and companies forget that the hiring process is a conversation between two parties. When you have a one-way recorded interview, I can’t ask them questions.” She, along with many people I spoke to, says it makes for a terrible first impression of a company. “If you don’t even take the effort to invite me to a company presentation, a meet-and-greet, or any sort of interaction before you throw me into your, you know, recruitment machine, then it’s a reflection of what your company culture is like.” Aurora, who hung up on a stilted A.I. interview with questions that made little sense, says this job hunt has been more demoralizing than her last. “The bot really upset me,” she says. “I feel like I need a job, but I’m just waiting for one to be an actual person that interviews me, so that I feel like I’m valued. I have a very clear idea of my worth, what I want to be paid, how I want to be treated—and that’s not how I want to be treated.” Lauren agreed: “It’s like you’re talking to a wall almost.” It’s no surprise that many of these automated hiring processes seem to be connected to the sorts of jobs that treat humans as machines anyway.
Parker acknowledges the lack of natural flow can be difficult but at least candidates have the opportunity to do retakes, unlike in real life. Plus, he adds, “we’re not looking for movie stars here. This isn’t a screen test.” (Jade, however, described it as feeling “like an audition reel.”)
This was, for me, the worst part of the mock interview HireVue sent me to try—an interview for a customer support representative with a made-up paper company. Although the questions were posed to me by cheerful fake staff from the company in prerecorded messages, I felt entirely alone—and I was, sitting in my bedroom trying to sell myself to A.I. Despite being given 30 seconds to prepare before answering each question, it was difficult to give full and engaging answers to no one, and my mind often went blank staring at either the timer or my own face. (This could be switched off, but I found the blank screen even worse.) What’s more, it was hard to quell my (very real) nerves without the niceties and introduction and small talk upfront.
That lack of chit-chat, Parker says, is actually one of the many advantages of these systems. “If we’re going to be hiring for you, everybody’s going to get the same experience, everybody’s going to get the same questions,” he says. It erases the advantage of some applicants being able to connect with an interviewer based on cheering for the same sports team or being from the same town, he says. There are other real benefits to A.I. interviews, too—as every single CEO I spoke to explained. Interviews can be done on demand, with no need for scheduling, meaning those currently working another job can interview without having to take time off. A typical Sunday for HireVue currently sees around 10,000 people interview, most on a mobile device. Many mentioned being able to make sure every candidate actually hears back, even the rejected ones. (Lauren says she never heard back from her HireVue, although it was just before the pandemic kicked off.) There is also an opportunity to provide automated, personalized feedback to every candidate—something PredictiveHire prides itself on. The company’s chatbot, Phai, is able to create and share constructive “insights” with a candidate almost immediately following an interview (“you take a logical and planned approach,” “you are independent minded”), along with “coaching tips” to work on particular attributes (“practice speaking your opinion”). CEO Barb Hyman wants to get this feature into the hands of students, and even those applying for jobs not being recruited by Phai.
And while many are concerned about these platforms’ ability to replicate existing bias, others are excited about the ability to overcome it—if done right. There’s one group that both HireVue and PredictiveHire mentioned as really enjoying their interviews when polled: those in the twilight years of their career.
After speaking to many of these A.I. firms, I believe that some of them do care about the candidate experience. HireVue has worked to improve its candidate experience based on the feedback of 1.5 million surveyed candidates. PredictiveHire’s “hiring with heart” sounds rather hollow, but Hyman and I spoke at great length about the need for humanity and empathy, especially now. “In a world of unemployment, how do you create some sort of dignity and some humanity when actually, life is pretty hard.” She said they avoided using video for this very reason.
But as much as these firms might try to ease the process for candidates, A.I. interviews exist because they are a cost-saving measure for recruiters. Viktor Nordmark, co-founder of Hubert+1 (with its “candidate experience fit for a king”), was rather blunt when I asked how he would respond to people who may feel undervalued by the process. “As companies you always seek to make more money,” he said. “I think it’s really hard to change that basic principle.”
The fact that these automated interviews are economically rational makes them no less upsetting for those who might be stuck doing them over the coming months and years. Maybe this is something we’ll all get used to, the way we’ve gotten used to Zoom; maybe HireVue’s smiling candidates really are having a fantastic time and are not just desperately trying to please the opaque A.I. But I keep thinking about a scene from the Netflix improv show Middleditch and Schwartz. When the comedians asked audience members for fodder from their everyday lives to kick things off, one offered up his A.I. job interview. Thomas Middleditch’s outraged horror said it all. “What the fuck is that?” he cried. “That’s so mean!”