Biased Algorithms?

http://www.nytimes.com/2015/07/10/upshot/when-algorithms-discriminate.html

In short, some professors suggest that algorithms should be made to be less discriminatory. Despite the fact that these algorithms are reflective of user data; these professors suggest that policy should be made to alter the algorithms to be more inclusive.

I won't give my opinion here in the first post, to be impartial. Discuss.

I entirely agree. Claiming "it's not our fault, it's just what came out of the algorithm" is passing the buck, and passing it to a nonexistent entity at that.

Human beings have mutually agreed to live by certain codes of conduct, and we don't get to dispose of them just because "durr I created a ROBOT that's doing the discrimination, not me, it's just math! How can math be discriminatory!!"

things a faulty algorithm can do:

* leak your private ssl key
* give access to your nude photos
* send all your bitcoins into the aether
* systematically identify black people as gorillas

guess which one is NO ONES FAULT REALLY

The New York Times wants me to login or whatever so I didn't read the article and I have no idea what algorithms we're talking about, or what kind of discrimination exactly. Like are they doing software for matchmaking or hiring people or whatever? Anyway if the "choose best candidate for the job" systematically favors skin color over skill level, the algorithm is A) written by a racist and B) getting access to data that it should not have access to, because it should be irrelevant.

Gez said:

The New York Times wants me to login or whatever so I didn't read the article and I have no idea what algorithms we're talking about, or what kind of discrimination exactly. Like are they doing software for matchmaking or hiring people or whatever? Anyway if the "choose best candidate for the job" systematically favors skin color over skill level, the algorithm is A) written by a racist and B) getting access to data that it should not have access to, because it should be irrelevant.

http://www.nytimes.com/2015/07/10/upshot/when-algorithms-discriminate.html

maybe this link will work

I willfully admit I thought this had to be something from TheOnion.

Then I actually did the incredible and began to read. Though upon completion, I now realize how many time's I've been racially profiled by captcha phrases, questions and pictures... assuming I know how to read, complete puzzles, and add numbers. Sitting there, taunting me, calling me out for never making it past the 3rd grade.

Pfff. How dare they judge me.

Is this about why when you type in the phrase "why are *ethnic race* people" something offensive pops up? I think Google fixed that a few years ago. Looking at it now its still got a few things, but not nearly the negatives it used to have.

I typed in "why are imps" and it auto filled...

Or is this about how many dogs show up in the Deep Dreaming?

I used to get ads for criminal pardons all the time. I wonder if my Irish name does it. Or perhaps the fact I'm an atheist Canadian? Who can say? (Facebook can say.) :D

I wonder what an algorithm that discriminates for/against Americans would look like. Names alone wouldn't work. Neither would location. Perhaps tracking a user's spelling and use of measurement units would work best.

Companies definitely need to pay attention to these things. They have incentive too. Excluding people from things they might click on is potentially cutting off some revenue.

A couple of weeks ago I was trying to create a US Battle.net account. But since I live in Europe, some region detection algorithm was very very insistent on redirecting me to the EU service no matter how much I manually selected the US region. In the end I needed someone from Blizzard support to create an account for me on their end and then hand it over to me.

That's one of the more infuriating cases of some annoying presumptious algorithm making things harder for me.

geo said:

Is this about why when you type in the phrase "why are *ethnic race* people" something offensive pops up? I think Google fixed that a few years ago. Looking at it now its still got a few things, but not nearly the negatives it used to have.

It's still pretty scary. It highlights common stereotypes, and almost every ethnicity is rude. According to google.ca:

• Canadians are nice.
  
• Asians are smart, short, yellow, and rude.
  
• Whites are rude, have thin lips, and are called crackers.
  
• Americans are stupid, rude, loud, and joining ISIS.
  
• Blacks are rude.
  
• Africans have yellow eyes and big lips.
  
• French are rude.
  
• Germans are rude.
  
• Polish are rude, made fun of, strong, and their guys are big.
  
• Chinese are loud, rich, bad drivers, and rude tourists.
  
• Japanese are short, weird, have funny eyes, bad teeth, and high voices.
  
• Koreans are pale (compared to whom?), and have hot guys.
  
• Australian guys are hot and rude.
  

There's two interesting examples in the article:

Google's online advertising system, for instance, showed an ad for high-income jobs to men much more often than it showed the ad to women, a new study by Carnegie Mellon University researchers found.

[...]

The Carnegie Mellon researchers who did that study built a tool to simulate Google users that started with no search history and then visited employment websites. Later, on a third-party news site, Google showed an ad for a career coaching service advertising “$200k+” executive positions 1,852 times to men and 318 times to women.

Research from the University of Washington found that a Google Images search for “CEO” produced 11 percent women, even though 27 percent of United States chief executives are women. (On a recent search, the first picture of a woman to appear, on the second page, was the CEO Barbie doll.) Image search results determined 7 percent of viewers' subsequent opinions about how many men or women worked in a field, it found.

I'd argue that the first example is a real problem -- the algorithm is evidently trained to distinguish between male and female users, and using it to discriminate. The second example, however, is harder to blame on the algorithm. Sure, 27% of CEOs in the USA are women, but does that mean that 27% of the pictures of CEOs on the Internet are women? "Deep Dreams" shows that Google is working on having some systems to automatically identify the content of every picture, so it might be possible to have the algorithm take "CEO" and conclude "display 50% of 'CEO' pictures showing men, and 50 of 'CEO' pictures showing women" but it doesn't seem to me that the system is in a working state right now. (Error: all CEOs are actually dogbirds in front of a pagoda.)

BUT LING, L0L:

Linguica said:

things a faulty algorithm can do:

* leak your private ssl key

If you're a good citizen and have nothing to hide, why would you need cryptography?

Linguica said:

* give access to your nude photos

A good citizen shouldn't have nude photos, of themselves or of others, except for medical purposes (not a good catch-all excuse unless you're a practicing MD, buddy).

Linguica said:

* send all your bitcoins into the aether

Bypassing the banking system, are we?

Linguica said:

* systematically identify black people as gorillas

Well, that was acceptable scientific practice well into the 19th century.

Linguica said:

guess which one is NO ONES FAULT REALLY

There are no faulty algorithms or laws, only faulty citizens, which WILL be fixed, sooner than later.

Aliotroph? said:

list

Well, at least nearly everyone is rude so that's not particularly discriminatory. ;)

Aliotroph? said:

• Africans have yellow eyes and big lips.
  

Not white teeth? I thought that was the most prominent feature (apart from the skin, ofc).

Is there a way to make Google not take any info about me into account? Just do a "normal" search that disregards my location, language, everything.

Memfis said:

Is there a way to make Google not take any info about me into account? Just do a "normal" search that disregards my location, language, everything.

Well, you could use one of those alternate "private" searches like Startpage or DuckDuckGo.

Memfis said:

Is there a way to make Google not take any info about me into account? Just do a "normal" search that disregards my location, language, everything.

You know, that raises an interesting aspect of all this:

First companies make it so they can collect user data for marketing purposes. This raises concerns of privacy of individuals being violated for profit.

Now professors are concerned that this same violation of privacy is too discriminatory. So instead, they want to change the out put of the collected data so people are treated equally.

May I propose this? How about making policy to not fucking collect search data for marketing purposes, or social regulation? Black, white, gay or straight; nobody wants to be annoyed by ads that are brought to you in such an intrusive way.

Maes said:

Not white teeth? I thought that was the most prominent feature (apart from the skin, ofc).

Oops. Forgot to add that one. It's definitely there. Belgians allegedly have ugly houses too.

Enough with this war on discrimination. It's literally the most important ability that all creatures possess. Discrimination has the wonderful property of rewarding those whose biases are correct, and punishing those whose biases are wrong.

To the extent that we negate discrimination, we negate our own success as beings.

Ads?

Memfis said:

Is there a way to make Google not take any info about me into account? Just do a "normal" search that disregards my location, language, everything.

Use Tor. Then it will conflate you with a vague group of anarchists, activists, journalists, pedos, and people interested in buying meth online.

Oh the horror of the discriminating algorithms and google showing you ; "are Americans... dumb ?". Have half of the posters complaining about both of those even read what they and many others are posting on DoomWorld ?

when it comes to math and humans i always come to the same conclusion.
As long as governments, courts, and cops stay away from any mathematician or tard claiming to have a mathematical model for humans or 'normal humans' the world should remain fine and dandy, anything or anybody else should shake hands with Hitler and the model for the normal human.

Google prioritizes sites that use adwords. I've done testing with dozens of sites having Google adwords vs not having them. You make them money, they'll make you money.

I've had to do a lot of research between Bing / Yahoo (same search provider) vs Google. Bing and Yahoo can seem like a randomized "who's on top." Keep refreshing and the order changes.

Google knows who you are and where you've been thanks to adwords and even Youtube since they're both everywhere.

That's why when I visit a site, I then see ads for it EVERYWHERE. Well wtf I went to a site without clicking an ad, now its ads are everywhere. I don't need the ad. I looked for a host for Doomworld (a day before it needed one), now I see the same ads of the sites I looked at. The algorithms aren't racist, sexist or bias, they're just based on where you've been.

If society stopped discriminating wouldn't the algorithms stop in return?

I'd argue that the first example is a real problem -- the algorithm is evidently trained to distinguish between male and female users, and using it to discriminate.

I don't get this. On one hand you usually rail against the evils of money and argue every executive is a psychopath. On the other hand an algorithm estimating women to be less driven by profit, more empathetic, more altruistic, and hence less interested in psychopathic work environments is, to you, a problem. Is the pursuit of equality and non-gender conformity these days so essential we even have to deny unequivocally positive characteristics in women?

That's one of the more infuriating cases of some annoying presumptious algorithm making things harder for me.

Now this is a real problem. Obviously it's not as juicy as complaining about the latest social media craze about racism or sexism (bless you Maes you crazy hermit greek, the stupidity of it all is made so obvious when someone outside of the noise speaks up), because Americans don't give a shit about foreigners being denied opportunities. Even the most liberal progressive will happily tell you Google, Apple, Amazon and so on do nothing wrong with their billions of tax avoidance, it's NO ONE FAULT'S REALLY as it's all legal and happens in one of your countries, you dirty europeans who are all the same.

Phml said:

Is the pursuit of equality and non-gender conformity these days so essential we even have to deny unequivocally positive characteristics in women?

http://blogs.scientificamerican.com/psysociety/benevolent-sexism/

Linguica said:

http://blogs.scientificamerican.com/psysociety/benevolent-sexism/

To conclude, I'll now ask you to think about recent events surrounding Elise Andrew, creator of the wildly popular I F--king Love Science Facebook page. When she shared her personal Twitter account with the page's 4.4 million fans, many commented on the link because they were absolutely SHOCKED...about what? Why, of course, about the fact that she is female.

"I had no idea that IFLS had such a beautiful face!"
"holy hell, youre a HOTTIE!"

"you mean you're a girl, AND you're beautiful? wow, i just liked science a lil bit more today ^^"

"I thought that because of all the ways you were so proud to spout off "I f--king love science" in a difient swary manner against people who hated sware words being used that you was a dude."

"you're a girl!? I always imagined you as a guy; don't know why; well, nice to see to how you look like i guess"

"What?!!? Gurlz don't like science! LOL Totally thought you were a dude."

"It's not just being a girl that's the surprise, but being a fit girl! (For any non-Brits, fit, in this context, means hot/bangable/shagtastic/attractive)."

Those guys must have some low-standards. But hey, that's totally subjective.



My rule of thumb is that if a girl gives you a smile, makes eye contact, and/or gives you specific signals as such; that's an invite to at least strike up a conversation. Women will make it abundantly clear that they're not interested in a man; usually by ignoring them. Thus cat-calling, and stuff like that is really a waste of time, and makes those guys look pathetic. Pathetic, as in the person has no idea how to find a woman that has a mutual-sexual desire for them.

Phml said:

Is the pursuit of equality and non-gender conformity these days so essential we even have to deny unequivocally positive characteristics in women?

As with mos things, depends on the context. If the discussion is framed as justification for discriminatory treatment, then we haven't solved the issue. It's throwing shade as a culture because while your evaluation of a person might be positive, the metric you use is negative. As some fundamentalist polygamist cultures would put it:

"You treat women like they're property."
"Yes, but INCREDIBLY VALUABLE property!"

Kontra Kommando said:

Those guys must have some low-standards. But hey, that's totally subjective.

It's well known that standards boil down to:

1. Breathes *
   
2. Has a vagina