I'm curious as to how this judges the skillset of a particular candidate. To me the only skill this would be able to judge is confidence, which also makes it very easy to trick - simply believe in your bullshit and you'll get through, no matter whether you actually have the skills.
Well, according to US political polls, 35% of the population is very happy with groupthink and tribalism despite any arguments. So there obviously is some room in business for a product like this.
35%? I think you underestimate the bad blood and ill faith in the US political system. Both major parties are mere parodies of genuine debate. I will concede that 35%, being Trump's approval rating, I assume, is an easy and valid lower bound.
Accessing this page results in an instant pop up box that says "Thanks for reading, like us on Facebook", even before you've had a chance to read a word in the article. What a fail. Won't read, pass.
This would, if allowed to run for generations, play out like incest - a dead end.
Fortunately for US-based companies, the quarterly earnings-per-share cycle creates so many decision changes within a year that a system like this wouldn't survive more than two years max. And that's about the tenure of the average employee, so it will wash out. Meanwhile, a boss will get a bonus for implementing it, and the vendor will make some cash. All is well, right?
To put skepticism aside for a second - I could see this working (maybe) in a field where face to face communication is incredibly important (e.g. sales). But the vast majority of roles in a modern company don't need people that are good at communicating verbally, not to mention in front of a camera.
Can you imagine putting a scientist through this process? A janitor? No way.
I think it's hard to avoid a lawsuit. An interview process that uses visual input to help it decide who to hire? It almost doesn't matter what else it considers - you'll have difficulty explaining to folk how it is not discriminating in some nefarious way.
Holy shit, what HR department would ever approve such a thing?! Photographs on resumes are discouraged - can you imagine the risk involved asking for a biometric scan and doing feature analysis on video interviews? How would this interact with GINA and polygraph laws? It raises far, far too many questions - unless the app makers are providing indemnity I can't see any reasonably competent hiring process professional adopting such a risky tool openly.
I remember Vodafone UK using this (at least the HireVue video interview thing, not sure if the AI-based approach described here was used). I turned it down straight away - if they can't bother allocating one hour of an interviewer's time for me then I shouldn't spend one of my own time for them.
This invariably will identify and rate characteristics like race. It will judge their vocabulary and identify if you are a native or an immigrant. Probably also identify your upbringing and class from language and grammar composition.
There are a thousand reasons why this is probably a garbage system. And chances are, it's just another AI-hype-driven thing to sell to ignorant companies/"decision-makers" who should be out of business already.
I always though that if the skills of a person are highly shaped by the genes, that would have a reflection in body traits. So basically you could have a good idea of the mental skills and personality of anyone just analyzing the body and other external features. One of the things that lead me to this theory is how mental illness do affect the body features.
But even if I am right in this theory, this only gives you one part of the person. Because the environment and the own decisions of a person shapes them. For example, even if a person if the perfect fit for the job position based on the analysis, and the analysis is right, the succeed of that person also depends in the attitude. If the person has developed a natural laziness over the life it wont be productive at the job. And the opposite can be true, a less genetical predisposed person which has work hard during all the life could be a much better match for the job.
EDIT: when I said the skills are encoded in genes and also expressed in body traits it doesn't mean gender or race, it could irrelevant things as the length of the fingers, or the way you walk.
This is only true if the two genes are correlated in success. I have faith such traits do exist, and many of them, but I am not immediately convinced most traits interesting to businesses would be revealed this way.
Also, your example of laziness is classic thinking yourself out of quality employees: laziness is good for identifying waste and inefficiency and for creating scaling automation. So it's not even clear if "laziness" is interesting without context to businesses.
Of course, it could also entail lack of work ethic.
Yeah but it affects a lot of other features (ie hearth defect). I don't really see distinctive features amongst bipolar or schizophrenic people... That seems like a rather generous extrapolation from one specific example
For me it is a theory that needs validation. I am not an expert in genetics but I think we still have a long way to know how the genes map to features.
> I don't really see distinctive features amongst bipolar or schizophrenic people...
What you don't see them doesn't mean they are not there. It could be something subtle. If one change in chromosome leads to a lot of mental and physical changes it could mean things like people with better maths skills grows faster the nails. And I repeat, this is a theory for me that still needs to be probed or at least more data. But I wouldn't discard it so fast. What this company is doing is going somehow in this direction. What do you think about the company of the article?
Some comments are focusing on this exacerbating aggregate phenomena like diversity stats and wage gaps. This is true and a huge negative.
But that's not the only thing we should be worried about. Far more damaging would be the dystopia that these "cognitive surveillance" products will bring upon us.
It claims to micro-analyze facial expressions, intonation, non-verbal signals while the candidate is interviewing. This is vile, hostile interaction in my opinion, and it is startling to see that companies like Vodafone, Intel, and Oracle are their customers [1].
At best, this is a way to sweep unaccountable decision-making under the carpet of "the software said so".
Quite likely, though, is that such products will make society a living hell for everyone until these practices are entrenched in the industry and it is too late to roll it all back.
The creators of these products are not stupid, they made a conscious choice to grab low-hanging fruit (now that they have the technology available) and enrich themselves while making the world a worse place. Let's not kid ourselves that the consequences of their actions did not occur to them.
As a research student in DL/AI, I realize this may make things marginally worse for my career, but right here is the reason we should regulate AI usage _now_: not Skynet, but these attempts to "disrupt" social norms for no stated reason other than "progress" and "efficiency".
We should keep some technologies out of the public sphere and make it vocally clear that they are unacceptable, lest we end up with a world where everyone is wearing google glasses and you have no way to maintain a personal facade because some combination of blood flow and facial muscle twitch betrays your thoughts.
Your only argument against this stuff should not only be that it has X bug or that it's execution isn't sound due to Y bias. We must oppose such things on principle, I would rather kill myself than live in some sort of Black Mirror-esque dystopia where this is mainstream.
At best, this is a way to sweep unaccountable decision-making under the carpet of "the software said so". Quite likely, though, is that such products will make society a living hell for everyone until these practices are entrenched in the industry and it is too late to roll it all back.
> It claims to micro-analyze facial expressions, intonation, non-verbal signals while the candidate is interviewing. This is vile, hostile interaction in my opinion.
Is your objection to the use of those things in general, or just to the use of an AI instead of a human to evaluate them?
I'm curious because these things are just, I think, the components that go into demeanor and humans routinely use demeanor to judge how much trust to put into what someone else is saying.
In fact, one of the main reasons that witnesses in a criminal trial in the US testify in person in front of the jury is so that the jury can see their demeanor and use it to judge credibility.
I'm pretty sure almost every human interviewer uses demeanor evidence, albeit not consciously. Short of eliminating face to face interviews I doubt that there is a way to stop such judging because it is almost completely unconscious.
I understand the idea of doing probablistic inference about people based on their demeanor. We do it all the time with our minds, indeed many aspects of society and culture are built around non-verbal communication.
I'm okay with how things currently are, because we are innately capable of learning behavior policies for social interactions, while preserving a lot of cognitive privacy. In other terms, you can consciously control your demeanor (as the word is currently understood).
I afraid of inferences drawn from sub-conscious demeanors and involuntary information leaks.
Normal humans will not be able to tell if someone is nervous or afraid or angry just by looking at them, if that someone wants to maintain a pokerface. But it is entirely possible to read someone's pulse by recording a video of their skin, thus taking away some privacy of the mind [1].
We haemorrhage a lot of information all the time. As far as I know, current polygraph tests are trash, but I would be unsurprised if some characteristic features were to be found in a video+audio stream of someone's face that would estimate with good accuracy and precision that they were lying.
We all lie. To others and to ourselves. Within limits, lying is an integral part of healthy life. A world where I cannot casually lie is not one where I can live in; cognitive privacy is important.
My understanding is that their objection is to having a black box AI that can make or break a hire. What's to stop them from not hiring anyone because "the box said no"? What's to stop the hiring process from being entirely composed of the black box, with nobody knowing what it takes to beat? Sure you have the skills and experience, but unfortunately your facial pattern makes you seem impersonal. Don't worry, they'll keep you on file just in case.
Wouldn't the free market solve for this? I mean if the AI system leads to worse hires the company hiring using it's signals will suffer in the market and ultimately lose out to a superior competitor who picked up hot superstar rejects with facial ticks on the cheap, right?
"Algorithms decide who gets a loan, who gets a job interview, who gets insurance and much more -- but they don't automatically make things fair. Mathematician and data scientist Cathy O'Neil coined a term for algorithms that are secret, important and harmful: "weapons of math destruction." Learn more about the hidden agendas behind the formulas."
...
"Algorithms are everywhere. They sort and separate the winners from the losers. The winners get the job or a good credit card offer. The losers don't even get an interview or they pay more for insurance. We're being scored with secret formulas that we don't understand that often don't have systems of appeal. That begs the question: What if the algorithms are wrong?"
These same algorithms run in our brains with far more flawed edge cases than big data. Not that I am defending blind use of "algorithms," but that word is a boogeyman at best.
That's why it's important to make sure people with significant variety in their backgrounds are involved in the process. Otherwise you run into situations where, for example, a team of designers doesn't discover their motion activated soap dispenser doesn't work on people with dark skin until after it's been produced, sold, and used.
We can talk about our decisions, explain them, and revise them if someone convinces us we're wrong. That's very different from a system that only outputs a vector of probabilities.
No, "these same algorithms" don't "run in our brains". Our brains produce thoughts in ways that we don't even have the slightest understanding of. We have conjecture and some observations (e.g. electric signal monitors) that only scratch the surface of describing what the end result of a "thought" looks like.
The comparison of how humans think and human intellect in general with poorly understood applied computational statistics really has gone too far.
Now, if you mean to say that humans themselves have biases and "wrong" behavior, that's a different matter.
This leads to some pretty scary questions. What about different cultures? Everyone has different ways of speaking, acting, living. What if the app considers something bad behavior, and it's just a cultural difference? Or it's how someone was raised?
Maybe the biggest question we should ask is what does this solve? What problem is this trying to solve? And once we figure that out, ask ourselves... is that the problem we should be solving? Is it the best way to solve it?
Is there any actual sound peer reviewed research indicating the results of their predictive system will lead to more great hires? Cus it didn't seem so in the article. The article made it seem like HireVue is working off some false causal relationship assumption where the best hires performed like this in their interview therefore we should compare all new hires to them without any shred of evidence showing that this type of interview performance indicates a good hire. I.e. snake oil...
Since when has science and reason driven management decisions? 20 years ago the silver bullet was OOP. 10 years ago the silver bullet was offshoring. Now it's AI and/or blockchain.
There's no way this software is not inherently biased on very basic measures--a lisp caused by a disability probably wouldn't pass muster, nor would a southern accent. But the real tragedy here is that it equates the way someone presents themselves--not the content of their words or the strength of their skills--with job performance. It exacerbates the worst vices of HR and hiring practices and performance reviews. It's an algorithm for superficiality.
This company should remove this product from the market immediately. How this passed off as a good idea all the way through the product development cycle is beyond me.
>This company should remove this product from the market immediately. How this passed off as a good idea all the way through the product development cycle is beyond me.
It sounds like you are assuming that businesses are supposed to act like a single human with ethics and a moral compass should. I agree with your conclusion but to ponder how this passed off as a good idea is silly. If I could come up with software to sell to businesses to replace a huge amount of their recruiting cost (people, time, operational cost) wouldn't I be inclined to do business with them?
Consider a company global company, when you look at hiring at scale it becomes easy to justify streamlining the whole process.
I still find it like a bad direction to go but what do you expect?
While there can be very much debate over the long term value of an AI based filter, at least the filter here includes much more than just "your face".
Honestly, I only clicked it to see if they were really just viewing a still image of your face and then trying to determine if you would be a good hire. I knew better.
The people funding the project with the fairly realistic expectation that they will cash out with a good profit one way or another within the next couple of years.
65 comments
[ 1.8 ms ] story [ 127 ms ] threadhttps://en.wikipedia.org/wiki/Phrenology
Fortunately for US-based companies, the quarterly earnings-per-share cycle creates so many decision changes within a year that a system like this wouldn't survive more than two years max. And that's about the tenure of the average employee, so it will wash out. Meanwhile, a boss will get a bonus for implementing it, and the vendor will make some cash. All is well, right?
0: Think of the plague
Can you imagine putting a scientist through this process? A janitor? No way.
> Goldman Sachs, Under Armour, Unilever, and Vodafone are also among the companies that have used the platform.
Let the lawsuits begin.
There are a thousand reasons why this is probably a garbage system. And chances are, it's just another AI-hype-driven thing to sell to ignorant companies/"decision-makers" who should be out of business already.
But even if I am right in this theory, this only gives you one part of the person. Because the environment and the own decisions of a person shapes them. For example, even if a person if the perfect fit for the job position based on the analysis, and the analysis is right, the succeed of that person also depends in the attitude. If the person has developed a natural laziness over the life it wont be productive at the job. And the opposite can be true, a less genetical predisposed person which has work hard during all the life could be a much better match for the job.
EDIT: when I said the skills are encoded in genes and also expressed in body traits it doesn't mean gender or race, it could irrelevant things as the length of the fingers, or the way you walk.
Also, your example of laziness is classic thinking yourself out of quality employees: laziness is good for identifying waste and inefficiency and for creating scaling automation. So it's not even clear if "laziness" is interesting without context to businesses.
Of course, it could also entail lack of work ethic.
Care to back up this claim?
> I don't really see distinctive features amongst bipolar or schizophrenic people...
What you don't see them doesn't mean they are not there. It could be something subtle. If one change in chromosome leads to a lot of mental and physical changes it could mean things like people with better maths skills grows faster the nails. And I repeat, this is a theory for me that still needs to be probed or at least more data. But I wouldn't discard it so fast. What this company is doing is going somehow in this direction. What do you think about the company of the article?
But that's not the only thing we should be worried about. Far more damaging would be the dystopia that these "cognitive surveillance" products will bring upon us.
It claims to micro-analyze facial expressions, intonation, non-verbal signals while the candidate is interviewing. This is vile, hostile interaction in my opinion, and it is startling to see that companies like Vodafone, Intel, and Oracle are their customers [1].
At best, this is a way to sweep unaccountable decision-making under the carpet of "the software said so". Quite likely, though, is that such products will make society a living hell for everyone until these practices are entrenched in the industry and it is too late to roll it all back.
The creators of these products are not stupid, they made a conscious choice to grab low-hanging fruit (now that they have the technology available) and enrich themselves while making the world a worse place. Let's not kid ourselves that the consequences of their actions did not occur to them.
As a research student in DL/AI, I realize this may make things marginally worse for my career, but right here is the reason we should regulate AI usage _now_: not Skynet, but these attempts to "disrupt" social norms for no stated reason other than "progress" and "efficiency". We should keep some technologies out of the public sphere and make it vocally clear that they are unacceptable, lest we end up with a world where everyone is wearing google glasses and you have no way to maintain a personal facade because some combination of blood flow and facial muscle twitch betrays your thoughts.
Your only argument against this stuff should not only be that it has X bug or that it's execution isn't sound due to Y bias. We must oppose such things on principle, I would rather kill myself than live in some sort of Black Mirror-esque dystopia where this is mainstream.
[1]: https://www.hirevue.com/customers
Sounds a lot like redlining to me[0][1].
[0] https://en.wikipedia.org/wiki/Redlining [1] https://www.nytimes.com/2017/08/24/upshot/how-redlinings-rac...
Is your objection to the use of those things in general, or just to the use of an AI instead of a human to evaluate them?
I'm curious because these things are just, I think, the components that go into demeanor and humans routinely use demeanor to judge how much trust to put into what someone else is saying.
In fact, one of the main reasons that witnesses in a criminal trial in the US testify in person in front of the jury is so that the jury can see their demeanor and use it to judge credibility.
I'm pretty sure almost every human interviewer uses demeanor evidence, albeit not consciously. Short of eliminating face to face interviews I doubt that there is a way to stop such judging because it is almost completely unconscious.
I understand the idea of doing probablistic inference about people based on their demeanor. We do it all the time with our minds, indeed many aspects of society and culture are built around non-verbal communication.
I'm okay with how things currently are, because we are innately capable of learning behavior policies for social interactions, while preserving a lot of cognitive privacy. In other terms, you can consciously control your demeanor (as the word is currently understood).
I afraid of inferences drawn from sub-conscious demeanors and involuntary information leaks.
Normal humans will not be able to tell if someone is nervous or afraid or angry just by looking at them, if that someone wants to maintain a pokerface. But it is entirely possible to read someone's pulse by recording a video of their skin, thus taking away some privacy of the mind [1].
We haemorrhage a lot of information all the time. As far as I know, current polygraph tests are trash, but I would be unsurprised if some characteristic features were to be found in a video+audio stream of someone's face that would estimate with good accuracy and precision that they were lying.
We all lie. To others and to ourselves. Within limits, lying is an integral part of healthy life. A world where I cannot casually lie is not one where I can live in; cognitive privacy is important.
[1]: http://people.csail.mit.edu/mrub/vidmag/
"The era of blind faith in big data must end" | TED Talk |
https://www.ted.com/talks/cathy_o_neil_the_era_of_blind_fait...
"Algorithms decide who gets a loan, who gets a job interview, who gets insurance and much more -- but they don't automatically make things fair. Mathematician and data scientist Cathy O'Neil coined a term for algorithms that are secret, important and harmful: "weapons of math destruction." Learn more about the hidden agendas behind the formulas."
...
"Algorithms are everywhere. They sort and separate the winners from the losers. The winners get the job or a good credit card offer. The losers don't even get an interview or they pay more for insurance. We're being scored with secret formulas that we don't understand that often don't have systems of appeal. That begs the question: What if the algorithms are wrong?"
The comparison of how humans think and human intellect in general with poorly understood applied computational statistics really has gone too far.
Now, if you mean to say that humans themselves have biases and "wrong" behavior, that's a different matter.
Maybe the biggest question we should ask is what does this solve? What problem is this trying to solve? And once we figure that out, ask ourselves... is that the problem we should be solving? Is it the best way to solve it?
This company should remove this product from the market immediately. How this passed off as a good idea all the way through the product development cycle is beyond me.
It sounds like you are assuming that businesses are supposed to act like a single human with ethics and a moral compass should. I agree with your conclusion but to ponder how this passed off as a good idea is silly. If I could come up with software to sell to businesses to replace a huge amount of their recruiting cost (people, time, operational cost) wouldn't I be inclined to do business with them?
Consider a company global company, when you look at hiring at scale it becomes easy to justify streamlining the whole process.
I still find it like a bad direction to go but what do you expect?
While there can be very much debate over the long term value of an AI based filter, at least the filter here includes much more than just "your face".
Honestly, I only clicked it to see if they were really just viewing a still image of your face and then trying to determine if you would be a good hire. I knew better.
Yes, because the prospective employer should be evaluating my job performance skills instead.