One-way video interviews are impersonal, candidates say, and raise privacy concerns

The increasing use of AI-enabled hiring tools often finds job applicants hyping their credentials in front of a reflection of themselves. Also known as “automated,” or “asynchronous” video interviews, one-way interviews have been marketed by some vendors as a way to cut through a deluge of applications with algorithms.

Some claim this is a common practice used by Fortune 500 companies. According to unpublished data shared with HR Brew, globally, "61% of recruiting and HR technology leaders say their organization is currently using [one-way video] technology," Jamie Kohn, a research director in Gartner’s HR practice, wrote via email.

Taking the short (video) cut

HireVue, Spark Hire and myInterview are among the providers of one-way video interviews—which typically involve a candidate filming themselves answering questions on a computer or phone, while a clock keeps track of an allotted time limit. Some applications use AI scoring metrics to assess candidate performance, but not all. Spark Hire, for example, provides video recordings sans AI, “for a number of reasons including the practicality/viability of AI models, the legal risk it could pose to our customers, and the candidate experience,” Matt Lerner, the company’s director of marketing, told HR Brew in an email.

Lerner wrote that HR departments can save time by relegating the scutwork of preliminary screening to a recording. “Let's say a recruiter has 100 applicants for a job and, on paper, 50 of them appear qualified. With a traditional phone screen, it just won't be feasible for that recruiter to speak with 50 people for 30 minutes at a time based on the pressure they face from the business to help hire for the open job.”

Some applicants, however, have complained online that the impersonal nature of the process can be daunting when trying to communicate one’s skills.

Capturing the real you

Sophie Hager, an industrial engineer who sat for two one-way video interviews over the course of a year, described difficulty “getting across your attitude, your personality, who you are, because you're really just focused on like, making sure that you're answering the question and not going too far out of what they asked for” during the process. “You’ve got a time limit [and] all of these restrictions that I think prevent you from really being able to fully answer the question the way that you would want,” she explained to HR Brew.

Depending on the specific technology being used, that glimpse of someone’s personality and aptitude might be processed by an algorithm that cobbles together a profile, using indirect cues—such as facial expressions, clothing, or background images—explained Ben Winters, counsel at the Electronic Privacy Information Center.

The “software might yield that you are not dependable or not good at X,Y, Z, or not trustworthy or whatever else, based off of some facial attribute, or something that you’re wearing, or something that’s in your screen,” Winters said.

Some technologies are different. For example, myInterview’s AI tech creates a transcript from its interviews, which is then “read by a machine-learning algorithm [which] scans and searches the content of the interview” for traits indicative of the “big five” personality traits, Clayton Donnelly, the company’s chief behavioral psychologist, explained. A team of psychologists working with myInterview watched over 30,000 interviews in an effort to identify what he described as indicators of the big five personality traits. “The machine-learning algorithm learns from what the psychologists do with the training set,” he said.

A recent paper by Ifeoma Ajunwa, an associate professor of law at the University of North Carolina’s Law School and the founding director of the artificial intelligence and decision-making research program, explores “AI-based video interview technology’s potentially discriminatory effects,” and job candidates who spoke to HR Brew said talking to flesh-and-blood recruiters provides an easier, less fraught experience.

Watching the clock tick

“The one-way interview is difficult… because you have no one that you’re talking to except the machine,” Ann John, an operations and logistics executive, told HR Brew. “You can’t probe the employer for additional information and clarity on the questions” being asked.

Typically, applicants are afforded limited time—about 60 seconds—to answer a question, according to a guide to automated interviews from Flexjobs. John said that she sat for four one-way interviews over six months, and all of them asked about five questions. In every instance, John said she wished she had more of an opportunity to impart her strengths through more personal means.

It’s “an unnatural environment to speak to somebody about something as important as a career choice…speaking into a machine just isn't the same,” she said.

Both Hager and John have made it past the automated-screening round, but researchers and privacy advocates warn that the AI employed by some of these tools could be rejecting candidates in a discriminatory manner.

Ajunwa wrote in her paper that “Automated technologies, which are often adopted as anti-bias interventions, have often been found to not only replicate the bias they were meant to evade but, in fact, also amplify it.”

Ajunwa’s paper mentions HireVue’s database of around 25,000 pieces of facial and linguistic information, which, she wrote, was used “to provide recruiters with a measure of a candidate’s potential job performance.” In 2021, following an internal audit, the company announced it had stopped using facial-analysis tech as part of its candidate screening assessments, but Ajunwa argued that its impact across the recruiting industry will endure: “HireVue’s marriage of AI and video interviewing has become entrenched as standard practice for recruitment,” she wrote.

“HireVue is committed to improving access to employment for a wider talent pool and reducing bias within the recruiting and hiring process. Our AI assessments integrate more than 100 years of research from the field of industrial organizational psychology with modern applications of digital technology and data science,” Amanda Hahn, HireVue’s VP of product marketing, said in a statement.

Not-so-hidden bias

Studies have indicated that algorithmic bias is most prevalent in the facial scans of non-white people: For example, a 2018 study by Lauren Rhue, assistant professor of information systems at the University of Maryland’s Robert H. Smith School of Business, found that two different facial-analysis programs determined that images of Black NBA players were respectively interpreted to be more angry or contemptuous than those of white players. (Of the two programs Rhue studied, Microsoft declined to comment and Megvii failed to respond to email requests for comment from HR Brew.)

Sophie Hager, the industrial engineer, said she’s now a part of hiring decisions at her current company, which also uses one-way video interviews. Though she’s now on the other side of the equation, some of the uneasiness she felt while interviewing as a candidate through the format still lingers: “As someone who's on the opposite side of it now where I'm grading people's responses and deciding whether or not they're a fit for the company, I think that I still feel the same exact way. I still feel uncomfortable watching it, knowing how those people feel.”