(This column originally appeared in Forbes)
If you’re wondering where AI is starting to become commonplace, look no further than your human resources department. I regularly see new applications and startups promising to help employers use AI to do better performance reviews, create job descriptions, manage paperwork and recruit talent.
HireVue has been in this game for more than a decade. Using this platform, businesses of all sizes can direct candidates to a website where interviews are done independently and on-demand using video and answering predetermined questions. The platform then prepares an assessment of the candidate based on the quality of their answers and submits those results to the recruiter who can then decide on next steps.
Addressing Bias
Biased? No. At least, not anymore.
In earlier iterations, the platform was evaluating behaviors like facial expressions, eye ticks, voice tone and body movements that was then reported back to employers. While doing only what a human interviewer would also be doing, its facial scanning and voice evaluation technology caused some to accuse the company of bias in its approach and lawsuits were filed. A few states, including California and Texas, banned the use of certain AI-features that could lead to bias in employee evaluations and hiring.
But the company learned, and a few years ago its executives focused on “resetting” the conversation and stressed how its new approach would make the job market “inclusive for all.” They removed those controversial features from their platform and now HireVue’s main focus is whether or not an employee is giving the right answers.
A Focus On Skills
Mike Hudy, HireVue’s Chief Science Officer, stresses that the application focuses on skills, not behavior, and is adamant that these features weren’t removed because of bias but simply because they didn’t improve the platform’s predictive accuracy.
“That’s all gone and our technology or science doesn’t even have access to that,” he said. “The data showed that complementing it doesn’t actually lead to incremental predictions, so we dropped it from the model.”
As a result, the company stresses that its system can actually reduce bias more effectively than humans. Hudy says that extensive adverse-impact testing is performed constantly and he finds its AI models to be more consistent and fairer than human evaluators.
“We are trying to help our customers tap into what skills and potential the individual has that aligns with the job,” he said. “When you focus just on skills, you start to level the playing field.”
So how is this done? By focusing on answers and predicting performance.
How It Works
The company has a proprietary scoring system that looks at interview responses strictly on language content that’s tied to job competencies. It uses a number of structured scoring “rubrics” (an assessment tool, often in a grid format, that explicitly outlines criteria, performance levels, and descriptors to evaluate performance) so recruiters can evaluate all candidates consistently.
Although its models are trained to replicate human evaluators AI helps the system get smarter with each interview. The intent is to replicate an expert evaluator and listen specifically for effective answers to the question posed. Because of this, any custom-written questions provided by employers are not included in the model so that its “scored” response analysis is only using HireVue-developed questions for consistency.
“When we’re using artificial intelligence to score, we are scoring only the content of the interview.” said Hudy. “Every candidate is scored on the same rubric, not on how they say something or how they look.”
HireVue, like many of its competitors, is geared towards companies that do high-volume hiring with their best-fit customers being large (5,000 to 10,000 employees), metric-driven employers. So even though it’s certainly applicable for smaller organizations, it may not be cost-justified in every case.
Hudy says that for most employers, the use of such platforms has become “pretty core.” And, thanks to its AI-based automation, it provides more opportunities both for the employer and for potential candidates.
More Opportunities For Employers and Candidates
“You can’t manually touch thousands of candidates,” he said. “We help to expand the aperture and allow more candidates to be considered.”
These platforms are still relatively new, and not every candidate may not be used to the process. Which is why HireVue requires its customers to make sure they’re transparent with their candidates from the start.
“There is a consent statement that AI is being used in the process and candidates can either opt in or opt out,” he said. “If they opt out, a human evaluates the interviewee using the same scoring rubric.”
He also advises to let candidates know that the platform will never be used to make hiring decisions and that there’s always a human involved. HireVue can make recommendations. But in the end it’s the recruiter’s call.
AI-driven recruiting tools like HireVue can have significant benefits. According to HireVue’s data, their clients report better hiring decisions assisted by its technology which leads to lower turnover and improved performance reviews. But we’re still in early days. Purdy says that future iterations of their platform will help their customers move resumes and job requisitions towards using AI to better match people to roles across an organization based on their skills.
“Resumes have always been garbage at predicting success,” he said. “Once AI is better used to match skills with roles we’ll see massive changes in the way organizations are managed.”
