patrick.net

A SJW Censor by any other name. Google has crested and will be on the information ash pile in a few years.

If giving a factually correct answer becomes a self-fulfilling prophecy, then it's reasonable to look at how to avoid this feedback effect.
For ex: if loans to black people are factually less likely to be repaid, should banks refuse all loans to all blacks based on no other information than the fact they are blacks?
It doesn't mean you need to lie or censor.

If something does not conform to my twisted delusions and mandates, are they by definition always algorithmic unfairness?

For ex: if loans to black people are factually less likely to be repaid, should banks refuse all loans to all blacks based on no other information than the fact they are blacks?

A bit of a straw man here. Blacks, as well as Whites, are individuals. Each individual qualifies or not based on that individual's employment record, earnings, and credit history. Individuals who do not score high enough in these measures tend to default. A better example of that self fulfilling policy is when banks are forced to use lower scores when approving certain applicants merely because those applicants are Black. This will of course result in a higher level of defaults for Blacks.

HeadSet says

For ex: if loans to black people are factually less likely to be repaid, should banks refuse all loans to all blacks based on no other information than the fact they are blacks?

A bit of a straw man here. Blacks, as well as Whites, are individuals. Each individual qualifies or not based on that individual's employment record, earnings, and credit history. Individuals who do not score high enough in these measures tend to default. A better example of that self fulfilling policy is when banks are forced to use lower scores when approving certain applicants merely because those applicants are Black. This will of course result in a higher level of defaults for Blacks.

Exactly right. And worse the banker shills and sunprime friends of mozilo peddled and pushed the loose lending standards to blacks so they can make a quick buck and leave the taxpayer with the default debt and virtue signal anybody out of the way who would rightly oppose this crap.

Google hung from its own SJW petard:

https://www.dailymail.co.uk/news/article-8029747/Googles-Thanksgiving-Four-interview-abruptly-fired.html

How predictable. Google plays Whack-a-Tranny.

Google a sellout lapdog of the Deep State? Whouda thought?

HeadSet says

Blacks, as well as Whites, are individuals. Each individual qualifies or not based on that individual's employment record, earnings, and credit history. Individuals who do not score high enough in these measures tend to default. A better example of that self fulfilling policy is when banks are forced to use lower scores when approving certain applicants merely because those applicants are Black. This will of course result in a higher level of defaults for Blacks.

The point is banks could deny loans to people only because they are black. It would make sense on a purely factual level. And this would tend to lock people out of opportunities requiring access to credit.
That's easy enough to understand.
And of course, instead banks should use individual characteristics.
And when in doubt, if you have to choose between losing a bit of money on a group or locking them out, then it's less harmful socially to default to the former.

Ceffer says

Google hung from its own SJW petard:

https://www.dailymail.co.uk/news/article-8029747/Googles-Thanksgiving-Four-interview-abruptly-fired.html

Leftists do-gooders who protest their own companies shouldn't work for US multinationals.
They're not even in the ballpark of reality.

Is it possible that Trans Women are mostly Mentally Ill?

Heraclitusstudent says

For ex: if loans to black people are factually less likely to be repaid, should banks refuse all loans to all blacks based on no other information than the fact they are blacks?

That is a truly terrible example. For the first reason, loans are already given based on INDIVIDUAL credit histories! Why would you want to apply a totally unfair GROUP criteria when individual histories offer a much better risk profile for each loan?

NoCoupForYou says

Is it possible that Trans Women are mostly Mentally Ill?

That might be a better option than regarding them as enlightened or oracular.

Heraclitusstudent says

And when in doubt, if you have to choose between losing a bit of money on a group or locking them out, then it's less harmful socially to default to the former.

When in doubt, don’t be a goddamned socialist nazi and group people by their fucking skin color when there’s INDIVIDUAL data available to work with! People are individuals. Some groups belong in the majority to cultures that can be reliably skewed towards different results. But applying a group cultural criteria (e.g. all whites are inbred honky stupid fucks) leads to an entirely INCORRECT assessment for a plurality of individuals. This is the definition of racism, and it’s combated with individualism.

Shaman says

That is a truly terrible example. For the first reason, loans are already given based on INDIVIDUAL credit histories! Why would you want to apply a totally unfair GROUP criteria when individual histories offer a much better risk profile for each loan?

Yeah... there are laws against discrimination.
But imagine you build a machine learning algorithm to decide who to give a loan to. The ML could notice blacks have problem repaying, and so decides based on this, or derived attributes (zip codes, first name, etc...). All of it is ENTIRELY factual. This is not like a bias introduced by the machine. This is a not a problem with the training set. This is a VALID generalization across individuals. Or it could be, and even if it was, you wouldn't want to use it.

And btw INDIVIDUAL credit scores are also based on GENERAL rules. And the color of your skin IS an INDIVIDUAL attribute. So what is generalized or individual? When you make decisions based on past history, you ALWAYS have implicitly general rules. Otherwise you couldn't extend past trends to new cases.

All this is why, this stuff is not so obvious as saying: "This is a fact so we can use it".
This is why there are people who are looking to explain ML decisions after training, which is not obvious in deep learning algos.

Yes, it is noble to work with people as iINDIVIDUALS so that they can fuck you over, destroy your business, and accuse you of sexism, racism, homophobia and all the other SJW canned aphorisms.

Selective vetting of employees and triage of potential client bases is how businesses survive. If it isn't your dime, obviously, it's a no-loss proposition to preach to them about social justice. This is called righteousness with no skin in the game. Hmm, is 'skin in the game' racis'?

Sounds like Google might be losing some skin to their pet revolutionaries and the class action wolf pack.

I wonder why it is deemed acceptable to charge males under 25 more money for car insurance than what females are charged?

rd6B says

I wonder why it is deemed acceptable to charge males under 25 more money for car insurance than what females are charged?

Interesting. So if I claim to "see myself" as female, do I get a rate cut?

Shaman says

Heraclitusstudent says

For ex: if loans to black people are factually less likely to be repaid, should banks refuse all loans to all blacks based on no other information than the fact they are blacks?

That is a truly terrible example. For the first reason, loans are already given based on INDIVIDUAL credit histories! Why would you want to apply a totally unfair GROUP criteria when individual histories offer a much better risk profile for each loan?

Agreed. Banks are the least racist organizations possible, caring only about money, and not at all about race.

The only color banks care about is green.

Heraclitusstudent says

The ML could notice blacks have problem repaying, and so decides based on this

The ML algo would only look at race if someone feeds it racial data and tells it to find relationships pertaining to that data.

You arent wrong though. Some relationships, like zipcode are seemingly more valid, but still out of control for many people. Zipcode being a rough model of wealth, and therefore credit worthiness...is that unfair or just a fact? ML deternined risk management rules such as "never loan money for a home in XYZ zipcode under XYZ price" is fair or biased? The answer isnt obvious, but I lean towards that being fair.

Heraclitusstudent says

This is a VALID generalization across individuals.

Auto insurance charges you more if you have a lower credit score. Is that fair? Or biased? They supposedly have data that suggests poor credit = more wrecks.

Guess which race has the worse credit.

HeadSet says

Interesting. So if I claim to "see myself" as female, do I get a rate cut?

Might work in progressive CA...

rd6B says

HeadSet says

Interesting. So if I claim to "see myself" as female, do I get a rate cut?

Might work in progressive CA...

Wonder if we will see a lawsuit where a 20 year old "pre-op" transwoman sues Geico for giving the transwoman the higher rate given to males.

Wait until the class action lawsuit, as all males identify as female to receive the extra money back they've paid over the last 6 years.

HeadSet says

rd6B says

HeadSet says

Interesting. So if I claim to "see myself" as female, do I get a rate cut?

Might work in progressive CA...

Wonder if we will see a lawsuit where a 20 year old "pre-op" transwoman sues Geico for giving the transwoman the higher rate given to males.

OR wait until life insurance carriers start rejecting trans-anything due to the 50% suicide rate, same way they are rejected from the military.

These are people in desperate need of help. Coddling their delusion is deeply irresponsible.