What Do We Know About AI-Powered Job Interviews?
The use of AI-powered technology in recruitment has skyrocketed over the past few years.
AI tools provide benefits like sourcing and screening thousands of candidates instantly, thus reducing the time and effort recruiters waste on tedious manual tasks.
At least that seems to be the general consensus.
As of 2019, AI is deeply embedded in recruitment, mostly through tools used in the earlier parts of the recruitment process – sourcing, screening and evaluating candidates based on their resumes and public information about them.
The next step in this evolution is clear – the interview process.
Job interviews led by AI aren’t that common yet, but more and more companies are implementing them, which means they must be seeing some positive results.
That said, we can’t hail AI interviews as the next best thing yet.
Giant corporations like Hilton, Under Armour, PWC, and Metro use their products.
While they might be different, these three tools offer the same benefits to recruiters – shorter and more convenient recruitment process, improved quality of hire, and a better ROI. One of their main selling points is the AI-powered job interview.
With these tools, interview questions are pre-recorded and presented to candidates who can then conduct the interview on their own time. As candidates answer the questions, their responses are being monitored and evaluated by AI.
These tools don’t evaluate just words, though.
They examine tonality, eye contact, micro-expressions, word clusters, and overall behavior to rate each candidate’s performance. Candidate performance then is measured against top performers in the company or against other candidates for the same job.
There are a few apparent benefits from conducting interviews this way. Candidates can answer the questions on their own time without worrying about formalities. Recruiters can examine responses based on candidate’s scores and focus their efforts on top performers. Thus AI provides both sides with a flexible solution that saves times and effort.
For example, VCV claim their product reduces the average recruitment process time from 21 hours to 45 minutes.
So far, so good.
While there are undeniable benefits from using AI to conduct interviews and evaluate result we still can’t assess a considerable number of crucial factors both technical and ethical.
First, let’s talk about the notion that AI reduces biases during the recruitment process. This is a commonly used selling point for companies that offer AI-powered tools.
For starters, AI is only as good as the data it’s fed. Saying that AI always reduces biases is like saying everyone starts with 100% “clean” and unbiased data. Which they don’t. Even giant companies like Amazon experience problems with bias AI during recruitment.
Tech companies are well aware of this problem. For example, IBM admits that in the next few years, the number of biased AI systems will only increase. IBM, Facebook, and Microsoft have all announced that they’re working on different solutions to the problem.
In a 2019 article titled “Notes from the AI frontier: Tackling bias in AI (and in humans)” the McKinsey Global Institute admitted that “AI can help reduce bias, but it can also bake in and scale bias.”The MIT Technology Review magazine has also published articles on why biased AI is so common and hard to fix.
So while huge tech companies and researchers are continually working on ways to improve AI, we can’t safely assume that AI assessment systems are entirely unbiased.
Secondly, the science behind some of these tools is murky, to say the least.
Most organizations that offer AI tools just slap an “approved by science” logo on their product and leave it at that.
For example, some companies state that their AI tool examines micro-expressions to evaluate candidates. Sounds cool on paper but in most cases, it’s just a marketing trick.
The topic of micro-expressions is a vast subject that can’t be covered in a single article or book. That said, most psychologists agree that the study of micro expressions is still in its infancy and most findings aren’t 100% reliable. There’s tons of conflicting research and lots of the early work on micro expressions (mainly developed by Professor Paul Ekman) is heavily critiqued by psychologists.
The final point we need to make concerns job candidates.
Mainly, do candidates want to be interviewed and evaluated by AI?
The use of AI in job interviews is still new, so there’s no hard evidence or comprehensive research on how job candidates in different industries feel about the use of AI. That said, we can look at a few indicators.
For example, if you look at reviews of HireVue on Facebook or other platforms, you’ll find lots of mixed opinions by job candidates. Some say it’s the most convenient interview experience ever while others are disgusted by the lack of humanity and transparency during the process.
Another question: Is it ethical for employers to evaluate candidates with AI without their knowledge?
In theory, collecting data about job candidates shouldn’t be a concern since, in most countries, employers are obligated by law to use the data only for the recruitment process and delete it afterward. That said, there has to be a limit to the information job candidates must disclose to get a job.
Right now, applying for most jobs means providing information about your education, skills, preferences, work experience, possible criminal history, etc. Employers state their requirements upfront, and candidates give this information willingly.
Things aren’t so transparent when using AI to conduct interviews and evaluate candidates. Since AI is mostly unregulated, there are no requirements forcing employers to notify candidates that they’re being assessed in such a way. That way candidates end up giving away lots of personal information unknowingly.
So, if these types of practices become the norm, job candidates will have to provide information about their background and let a machine learn how they talk, react to specific questions, where they look, when they smile, what tonality they use, and an array of other highly specific psychological traits.
Again, this data should only be used by the employer for the recruitment process. This doesn’t make collecting it without the candidate’s consent any less troubling.
Finally, if candidates are made aware of their evaluation via AI, can they refuse and if so, can employers reject them on this basis alone?
Of course, employers have the right to choose who they interview. That’s non-negotiable. At the same time, rejecting candidates just because they don’t want to give so much personal information doesn’t seem like a fair solution.
The dawn of AI is here and we need to make some crucial decisions.
Recruitment is yet another field where more and more companies are implementing machine learning.
It’s still too early to tell, but right now it seems like there are more questions than answers when it comes implementing AI in job interviews.
Companies might be seeing some benefits, but candidates must also take a stance on the matter.
If they don’t, they might be left in a situation where they must disclose everything about themselves to get a job.
If this all feels like you’re reading a George Orwell novel or watching Blade Runner…well it kind of is.