I was part of a very in-depth evaluations of both Pymetrics & HireVue for a large company recently (2 different evals, among others). Some interesting things we observed:
(1) It was not clear to us, with any level of certainty, whether either tool did anything more useful than filter out people who don't want to take the test / don't trust "AI".
(2) It was not clear to us, with any level of certainty, whether (1) was a useful, arbitrary, or counterproductive selection criterion for the roles we were hiring for.
(3) It was very clear to us, with a high degree of certainty, that both impressed a LOT of the higher ups
(4) It seemed somewhat clear that both were unlikely to get us sued (note: this is a high bar). However, HireVue (as of a year ago) could not in ANY meaningful way substantiate this claim "AI is more impartial than a human interviewer, as it has no bias," and it angers me that they make that claim.
After conducting a bunch of reviews like this (including other tools), I concluded that most AI recruiting software is a combo of (a) very useful process automation and (b) hocus-pocus, magicical, pseudo-AI. I had people telling me they used supervised learning because they had a team of supervisors in India. I had people telling me they had "Custom AI" when they were just calling some random unverified third party API. And I OFTEN heard nonsense, unverified claims of "unbiased" AI, as if eg training a hiring model using geographical factors won't tell you "don't hire non-college-age people from the south side of Chicago".
If anyone is looking to use recruiting AI, get a good IO psych person and listen to them.
(1) It was not clear to us, with any level of certainty, whether either tool did anything more useful than filter out people who don't want to take the test / don't trust "AI".
(2) It was not clear to us, with any level of certainty, whether (1) was a useful, arbitrary, or counterproductive selection criterion for the roles we were hiring for.
(3) It was very clear to us, with a high degree of certainty, that both impressed a LOT of the higher ups
(4) It seemed somewhat clear that both were unlikely to get us sued (note: this is a high bar). However, HireVue (as of a year ago) could not in ANY meaningful way substantiate this claim "AI is more impartial than a human interviewer, as it has no bias," and it angers me that they make that claim.
After conducting a bunch of reviews like this (including other tools), I concluded that most AI recruiting software is a combo of (a) very useful process automation and (b) hocus-pocus, magicical, pseudo-AI. I had people telling me they used supervised learning because they had a team of supervisors in India. I had people telling me they had "Custom AI" when they were just calling some random unverified third party API. And I OFTEN heard nonsense, unverified claims of "unbiased" AI, as if eg training a hiring model using geographical factors won't tell you "don't hire non-college-age people from the south side of Chicago".
If anyone is looking to use recruiting AI, get a good IO psych person and listen to them.