I think more scary than any bias the algorithm might have, is giving such enormous power to a single company - to be able to influence hiring decisions for all the companies that use it. Even if they're not malicious (yet), what happens when its use becomes more widespread, and some people are (rightly or wrongly) classified as unemployable? Right now they can try their luck at a different company, but it won't make any difference if that company also uses HireVue.
Putting a black-box AI in charge of everyone's prosperity is crazy.
I seriously have doubts about it becoming widespread.
It reminds me of the Meyers Briggs personality test that was also snake oil that got pushed onto companies. Pretty soon, people figured out how to game the test to guarantee the best possible outcome. If this thing ever goes bigger, people will do the same, especially with tech already existing.
And of course, no self respecting company will ever use this.
well, it only took us a hundred years or so but looks like we're back to craniometry, AI edition.
Also, a cursory look at HireVue's website shows an advertisement as follows: "Screen the best tech talent, no need to understand the code", enough said.
If anyone who works at that company reads this: You are making the world a worse place, you're selling people snake-oil, and you're building software that controls people rather than liberates them, spend a few minutes pondering the ethics of the systems you're building.
Id like to offer slightly more objective view on this:
I don't think anyone is going to disagree that this machine learning model doesn't work for selecting people that up the revenue of the company compared to someone it doesn't select. So there are 3 major uses cases for it from my perspective:
1. Established and proven tech companies understand that this is bogus, and choose not to use this. If they are looking to improve their hiring efficiency, they will most definitely do a review of this, and figure out that this is bogus.
2. Newer companies falsely believe in this tech and choose to use it - these are in the same boat as any other company that practices "bespoke" selection critera, i.e not hiring you because you wore khakis and polo to the interview for a non customer facing position instead of a suit, when in reality you may be the most qualified to do the job. You can't really criticize these, and Id argue that this is more of a loss to the company than yourself, as you would probably be better off at a different company anyways from overall long term money value perspective.
3. Bigger companies use this because of a special contract with HireVue to boost HireVue visibility or some available money to "try new things" to staff a few positions, most likely lower level that are easily replaceable so that if the program fails its an acceptable monetary loss. Here lies the biggest unfairness, due to the fact that people who are hired with HireVue's high scores are probably just as equally capable of doing the job as people who don't get hired, which at the minimum reinforces the belief that the model is working. That being said, these selection tactics are not new, and unemployment in US is low and still predicted to fall, and these jobs aren't career changing, so I would argue that this is not a huge issue either.
So my personal opinion as a capitalist is that this isn't a big deal. In the same way that you never see the Meyers Briggs Personality test outside of lower level employment, you won't see this for any job that can determine your career. And, just like Meyers Briggs, there will be ways to game it. Hell, the tech even exists already (a la Deepfakes), all you need is a rough estimate of the model they use and what the positive traits are, and then you can do real time face and voice morphing.
Its also generally more beneficial to think about adaptation rather than control or blame when it comes to stuff like this, because we can't un-invent the processing power and research makes machine learning possible at a scale. Someone could have made this open source, and then companies could internally implement their own tech stacks for video interviews, and we would be in the same boat.
> “I feel like that’s maybe one of the reasons I didn’t get it: I spoke a little too naturally. Maybe I didn’t use enough big, fancy words. I used ‘conglomerate’ one time.”
One of the bad outcomes will be if people imagine ways of gaming the system (like using big words, or smiling broadly), and then cargo-cult those into real life.
People are already prone to this, especially when applying for jobs with vague criteria for success like investment banking.
I think if you told people they're being judged by an AI, and then simply decided randomly, they would cook up no end of stories about how it works and what it's looking for.
Voight-Kampff style snake oil, wrapped in AI phrenology. As these guys don't provide feedback to the test subject and have not had their tech audited, who knows what this is based on and how mature the technology is. And major firms are basing hiring decisions around this? Unbelievable.
9 comments
[ 3.2 ms ] story [ 32.5 ms ] threadPutting a black-box AI in charge of everyone's prosperity is crazy.
It reminds me of the Meyers Briggs personality test that was also snake oil that got pushed onto companies. Pretty soon, people figured out how to game the test to guarantee the best possible outcome. If this thing ever goes bigger, people will do the same, especially with tech already existing.
And of course, no self respecting company will ever use this.
Also, a cursory look at HireVue's website shows an advertisement as follows: "Screen the best tech talent, no need to understand the code", enough said.
If anyone who works at that company reads this: You are making the world a worse place, you're selling people snake-oil, and you're building software that controls people rather than liberates them, spend a few minutes pondering the ethics of the systems you're building.
Ecclesiastes 1:9
I don't think anyone is going to disagree that this machine learning model doesn't work for selecting people that up the revenue of the company compared to someone it doesn't select. So there are 3 major uses cases for it from my perspective:
1. Established and proven tech companies understand that this is bogus, and choose not to use this. If they are looking to improve their hiring efficiency, they will most definitely do a review of this, and figure out that this is bogus.
2. Newer companies falsely believe in this tech and choose to use it - these are in the same boat as any other company that practices "bespoke" selection critera, i.e not hiring you because you wore khakis and polo to the interview for a non customer facing position instead of a suit, when in reality you may be the most qualified to do the job. You can't really criticize these, and Id argue that this is more of a loss to the company than yourself, as you would probably be better off at a different company anyways from overall long term money value perspective.
3. Bigger companies use this because of a special contract with HireVue to boost HireVue visibility or some available money to "try new things" to staff a few positions, most likely lower level that are easily replaceable so that if the program fails its an acceptable monetary loss. Here lies the biggest unfairness, due to the fact that people who are hired with HireVue's high scores are probably just as equally capable of doing the job as people who don't get hired, which at the minimum reinforces the belief that the model is working. That being said, these selection tactics are not new, and unemployment in US is low and still predicted to fall, and these jobs aren't career changing, so I would argue that this is not a huge issue either.
So my personal opinion as a capitalist is that this isn't a big deal. In the same way that you never see the Meyers Briggs Personality test outside of lower level employment, you won't see this for any job that can determine your career. And, just like Meyers Briggs, there will be ways to game it. Hell, the tech even exists already (a la Deepfakes), all you need is a rough estimate of the model they use and what the positive traits are, and then you can do real time face and voice morphing.
Its also generally more beneficial to think about adaptation rather than control or blame when it comes to stuff like this, because we can't un-invent the processing power and research makes machine learning possible at a scale. Someone could have made this open source, and then companies could internally implement their own tech stacks for video interviews, and we would be in the same boat.
One of the bad outcomes will be if people imagine ways of gaming the system (like using big words, or smiling broadly), and then cargo-cult those into real life.
People are already prone to this, especially when applying for jobs with vague criteria for success like investment banking.
I think if you told people they're being judged by an AI, and then simply decided randomly, they would cook up no end of stories about how it works and what it's looking for.