4 ms·

I'm not sure if there was anything deliberate here. It's possible the companies targeted the jobs ads with factors or attributes that were correlated with young

I'm not sure if there was anything deliberate here. It's possible the companies targeted the jobs ads with factors or attributes that were correlated with younger men by the Facebook algorithm.

deogeo 7y ago
Being deliberate is not required: https://en.wikipedia.org/wiki/Disparate_impact https://en.wikipedia.org/wiki/Disparate_impact
- mlthoughts2018 7y ago
  > “Where a disparate impact is shown, the plaintiff can prevail without the necessity of showing intentional discrimination unless the defendant employer demonstrates that the practice or policy in question has a demonstrable relationship to the requirements of the job in question.[3] This is the "business necessity" defense.[1]” I wonder if you could argue that advertising in a particular way is a business necessity because without doing so, the business cannot maintain competitive viability in reaching necessary candidates. Just for a devil’s advocate example, suppose that travelling to university recruiting fairs to recruit for computer science job listings has a disparate impact against female candidates, because the gender composition of university programs in computer science are heavily skewed towards male applicants. But university program graduates may be the only qualified type of candidate for the position, so simply not targeting candidates in the male-overrepresented way (i.e. not attending these university job fairs) would be debilitating and hugely damaging for employers who require those qualified candidates to be competitive in the market.
  - yawaramin 7y ago
    Yes, the hypothetical you described is one potential scenario of the line you quoted: > ... unless the defendant employer demonstrates that the practice or policy in question has a demonstrable relationship to the requirements of the job in question.[3] This is the "business necessity" defense.[1]” But it's still just a hypothetical. The only way to get more certainty is for you to hire a lawyer to advise you on the matter. The only way to get absolute certainty is to actually carry out the scenario and go through a lawsuit :-)
  - shkkmo 7y ago
    You can have an employment policy that favors CS grads when hiring programmers because, while the policy is effectively discriminatory, you can show that it directly relates to the job being hired for.
    - mlthoughts2018 7y ago
      It seems this would easily extend to many more types of job recruitment targeting then too, no? Or at least targeting job ads seems like it would be done for the exact same reasons as favoring CS grads in recruitment, until clearly proven otherwise?
- ConradKilroy 7y ago
  As a wikipedian, wow thank you, I love learning vital descriptive terms.
- macspoofing 7y ago
  Sure, but it changes the optics from those companies doing actively something nefarious, to doing it accidentally out of ignorance of the underlying technologies.
uberman 7y ago
As the test case I feel it is rather unlikely. One typically goes after the sure offender.
mlthoughts2018 7y ago
It’s interesting to imagine a hypothetical. Suppose a company chose an ad targeting segment with perfect 50/50 historical probability of a viewer being of either traditional male / female gender identity given they are in the segment. Now suppose through some weird fluke, the ads are actually only delivered to people of one single gender classification. Has the company done everything required of them by trying to target an equal mix along this particular characteristic dimension that society attaches a legal criteria to? Or has the company failed to do what is required because they merely tried to target 50/50 but failed to guarantee that it actually happened? Then flip it around. Suppose a company targets an ad segment that is 100% male because they are actively discriminating against women. But through some fluke of ad delivery they actually show the ad to a 50/50 mix. Are they free of committing a crime because nobody ended up being discriminated against? Or have they committed a crime merely by trying to target a certain subset of the population with a disapproved set of characteristics? To me it begs a question of whether the company’s legal obligation occurs at the level of defining who to try to target, at the level of who is actually targeted (even if it is executed by a third party), or both? A further interesting question is that if this applies to one type of resource (jobs) does it also apply to others like plain old goods and services? Suppose a sporting goods company knows that only 100% males have ever bought their merchandise or expresssed willingness to buy in market research. Is the company legally obligated to still show ads to women? If not, what is the underlying criteria that supposedly separates job listings from product listings? Another interesting question is whether accidental correlation with mix of the targeted audience also results in blame. What if you target a set of people with behavioral characteristics A, B and C. None of them has any obvious connection to gender, sexual orientation, etc., but it turns out they are heavily correlated and (honestly) nobody knew ahead of time (maybe even the correlation is seasonal and prior data could not have shown it). Does this make the company responsible?
- toss1 7y ago
  For your examples of intent vs actual net targeting, it is not unreasonable to expect that an employer check the actual results of their targeting selections and ensure that they produce sound results. "what is the underlying criteria that supposedly separates job listings from product listings?" It is called Equal Opportunity Employment Law. If you want to employ someone, you MUST give equal opportunity, and cannot discriminate based on any of the protected classes (and ignorance of the law is no excuse). Re your various characteristics, again, you MUST check your results for parity of results. Your example of "oh, I'm just targeting these particular characteristics and I had no idea that they happened to correlate 100% with young white males, oh my!", would provide such an easy work-aroud the EEO laws as to make them worthless. So, no, you can't do that.

I'm not sure if there was anything deliberate here. It's possible the companies targeted the jobs ads with factors or attributes that were correlated with younger men by the Facebook algorithm.

deogeo 7y ago

Being deliberate is not required: https://en.wikipedia.org/wiki/Disparate_impact https://en.wikipedia.org/wiki/Disparate_impact

mlthoughts2018 7y ago
> “Where a disparate impact is shown, the plaintiff can prevail without the necessity of showing intentional discrimination unless the defendant employer demonstrates that the practice or policy in question has a demonstrable relationship to the requirements of the job in question.[3] This is the "business necessity" defense.[1]” I wonder if you could argue that advertising in a particular way is a business necessity because without doing so, the business cannot maintain competitive viability in reaching necessary candidates. Just for a devil’s advocate example, suppose that travelling to university recruiting fairs to recruit for computer science job listings has a disparate impact against female candidates, because the gender composition of university programs in computer science are heavily skewed towards male applicants. But university program graduates may be the only qualified type of candidate for the position, so simply not targeting candidates in the male-overrepresented way (i.e. not attending these university job fairs) would be debilitating and hugely damaging for employers who require those qualified candidates to be competitive in the market.
- yawaramin 7y ago
  Yes, the hypothetical you described is one potential scenario of the line you quoted: > ... unless the defendant employer demonstrates that the practice or policy in question has a demonstrable relationship to the requirements of the job in question.[3] This is the "business necessity" defense.[1]” But it's still just a hypothetical. The only way to get more certainty is for you to hire a lawyer to advise you on the matter. The only way to get absolute certainty is to actually carry out the scenario and go through a lawsuit :-)
- shkkmo 7y ago
  You can have an employment policy that favors CS grads when hiring programmers because, while the policy is effectively discriminatory, you can show that it directly relates to the job being hired for.
  - mlthoughts2018 7y ago
    It seems this would easily extend to many more types of job recruitment targeting then too, no? Or at least targeting job ads seems like it would be done for the exact same reasons as favoring CS grads in recruitment, until clearly proven otherwise?
ConradKilroy 7y ago
As a wikipedian, wow thank you, I love learning vital descriptive terms.
macspoofing 7y ago
Sure, but it changes the optics from those companies doing actively something nefarious, to doing it accidentally out of ignorance of the underlying technologies.

uberman 7y ago

As the test case I feel it is rather unlikely. One typically goes after the sure offender.

mlthoughts2018 7y ago

It’s interesting to imagine a hypothetical. Suppose a company chose an ad targeting segment with perfect 50/50 historical probability of a viewer being of either traditional male / female gender identity given they are in the segment. Now suppose through some weird fluke, the ads are actually only delivered to people of one single gender classification. Has the company done everything required of them by trying to target an equal mix along this particular characteristic dimension that society attaches a legal criteria to? Or has the company failed to do what is required because they merely tried to target 50/50 but failed to guarantee that it actually happened? Then flip it around. Suppose a company targets an ad segment that is 100% male because they are actively discriminating against women. But through some fluke of ad delivery they actually show the ad to a 50/50 mix. Are they free of committing a crime because nobody ended up being discriminated against? Or have they committed a crime merely by trying to target a certain subset of the population with a disapproved set of characteristics? To me it begs a question of whether the company’s legal obligation occurs at the level of defining who to try to target, at the level of who is actually targeted (even if it is executed by a third party), or both? A further interesting question is that if this applies to one type of resource (jobs) does it also apply to others like plain old goods and services? Suppose a sporting goods company knows that only 100% males have ever bought their merchandise or expresssed willingness to buy in market research. Is the company legally obligated to still show ads to women? If not, what is the underlying criteria that supposedly separates job listings from product listings? Another interesting question is whether accidental correlation with mix of the targeted audience also results in blame. What if you target a set of people with behavioral characteristics A, B and C. None of them has any obvious connection to gender, sexual orientation, etc., but it turns out they are heavily correlated and (honestly) nobody knew ahead of time (maybe even the correlation is seasonal and prior data could not have shown it). Does this make the company responsible?

toss1 7y ago
For your examples of intent vs actual net targeting, it is not unreasonable to expect that an employer check the actual results of their targeting selections and ensure that they produce sound results. "what is the underlying criteria that supposedly separates job listings from product listings?" It is called Equal Opportunity Employment Law. If you want to employ someone, you MUST give equal opportunity, and cannot discriminate based on any of the protected classes (and ignorance of the law is no excuse). Re your various characteristics, again, you MUST check your results for parity of results. Your example of "oh, I'm just targeting these particular characteristics and I had no idea that they happened to correlate 100% with young white males, oh my!", would provide such an easy work-aroud the EEO laws as to make them worthless. So, no, you can't do that.