14 ms·
> the dataset contained ... demographic attributes like race ... to predict actual work performance Regardless of the original intent, that's a request to laun
by pdkl95 7y ago
> the dataset contained ... demographic attributes like race ... to predict actual work performance
Regardless of the original intent, that's a request to launder a protected class into an opaque "work performance" metric. Also, while they are not on the unfortunately short list of protected classes, using demographics like age is probably going to create similar discrimination problems.
If you want to know about job performance, use data that is at least theoretically related to someone's performance. While machine learning is going to find some sort opf signal in whatever pile of data you feed it, that doesn't mean it actually predicts something complicated like job performance. When you feed it a bunch of general demographics, it's going to find one of the signals that we know are present throughout existing demographic: the institutional discrimination and similar problems of the past.
- ergothus 7y agoExcellent point! Isn't age a protected class though? At least in the US, discrimination over 40 is covered by the EEOC.
- roflc0ptic 7y agoI'm hazy on the details, but it's not a protected class in the same way that e.g. race is. In alleging racial discrimination (in certain legal venues), statistical evidence of racial discrimination is admissible, however statistical evidence of age discrimination is not. E.g. if there are layoffs, and every grey head disappears, that's not grounds to allege discrimination. You'd have to have someone on record saying "we're firing all of the old fucks because they remind me of the inescapability of my own mortality" to litigate. Someone with a better grasp on the particulars here please chime in :)
- gumby 7y agoThey are protected against certain kinds of discrimination, but no human discriminated against them. It was simply a dispassionate algorithm and an algorithm can't display malice! Grrr... or that's what they, and probably the courts, would say. The OP calling it "discrimination laundering" had it right.
- pkaye 7y agoI don't think you can use "no human discriminated against them" as a defense. It could still be constructive discrimination.
- lmkg 7y ago> probably the courts, would say The courts are the ones that created the doctrine of "disparate impact." A criterion can be discriminatory even if it doesn't look discriminatory on its face. And, importantly, it does not require intent to discriminate, only impact.
- sterkekoffie 7y agoI actually thought that line was a joke when I first read it. "We don't hire non-citizens, Hispanic people, or anyone over 30" sounds a lot worse than "we evaluate candidates according to an algorithm trained to predict work performance."
- cannonedhamster 7y agoBefore tech I did a stint in recruiting. We had a client demand only "white sounding" names. We ignored that and provided any qualified candidate. They complained because they mistook the recruiters name for an applicant. We fired that client the same week. It was a truly disgusting thing to be asked and I'm glad my company made the right call. Any future client that asked that was immediate grounds for client dismissal, i.e. didn't require management approval.
- RobertDeNiro 7y agoI'm conflicted about this, on the one hand it's completely wrong, but on the other hand would a non-white person even want to work for them? In the end their own biases will cause them to fail.
- shims 7y agoThere are plenty of people who need to work to eat and can't be picky with employers.
- giggles_giggles 7y agoI spent the whole time reading the article thinking "so you got scammed by people looking to find a way to hide prejudice behind ML? What a surprise..." Maybe the lesson here is that people doing unethical things will take the opportunity to do more unethical things, like not pay you, and it may well be in your self interest to apply your ethics when choosing a job? Upwork doesn't look great here, but seems like it's hardly the whole problem.
- hihi123 7y agoNo, that's not the lesson. Upwork the platform scammed him, not the 'unethical' clients. The clients were looking to settle the bill.
- oeoeo00 7y agoThe client says that but they’re peddling unethical shit. It’s not unlikely the client knows exactly the way Upwork works and gamed their process too. And they get away with by scapegoating Upwork. Ticketmaster is an abuse catcher for venue price gouging. Why not hide behind Upwork policy to achieve a similar “no skin in the game” buffer. We just keep adding administration layers to everything, for jobs of course, while we see it always turn into an obfuscation tactic for swindlers.
- closetohome 7y agoI mean, the clients may be unethical too. But more in a not-hiring-minorities kind of way than a scamming-their-contractors kind of way.
- bofadeez 7y agoThat's a pretty wild assumption. They clearly were willing to pay.
- opwieurposiu 7y agoIf you think it is not possible for race be related to job performance you must not have watched much NBA basketball.
- jklinger410 7y agoes ist verboten "bad science"
- gurkendoktor 7y agoOr Sherpa people carrying luggage up Mt Everest... There's a difference between acknowledging some correlations, and using them as statistical tools in the job market. The latter is thankfully illegal or at least frowned upon.
- DocTomoe 7y agoWhen they built New York's skyscrapers, they hired predominantly people from the Mohawk tribe to do riveting a three hundred feet above ground because they lacked fear of heights. At what point does a correlation become useful as a statistical tool?
- gurkendoktor 7y agoEvery proven correlation is immediately useful as a statistical tool! But we also don't jail people when there's an 80% chance they're guilty. Not everything is about efficiency.
- 3131s 7y agoThe best athletes are extremely intelligent and driven people. The over-representation of black people in athletics is most definitely related to their under-representation in other fields.
- MuffinFlavored 7y agoI think it is less "intelligence" and more "naturally physically talented"
- sct202 7y agoHaving race in the dataset itself might not be bad, if there was an expectation that the data scientist would use the data to check if the algorithm was biased in some way, but it seems like the customer was pretty terrible sooo who knows.
- j05huaNathaniel 7y agoWhat if this was the NBA? Are you telling me that race wouldn't be an indicator of work performance? Are you telling me that age wouldn't be an indicator of work performance? You're also assuming some malicious intent when you have zero context. Maybe stick to just talking about the topic at hand.
- bofadeez 7y agoComments like this make me cringe so hard. People like you are why evolutionary biologists have to do their work in secrecy. You're basically a Luddite with respect to psychometric/statistical tech available now via identical twin studies. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5068715/ https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5068715/ (nobody will reply, everyone will downvote. this is a result of their cortex being overwhelmed.)