Hacker News Reader

Comment by adrianN | Hacker News Reader

The accuracy or inaccuracy of stereotypes reminds me of the regulations and ethical debates regarding the use of machine learning for example for creditworthiness. The algorithms often turn out to become "racist", because from the data they have, race is a good predictor.

I'm still somewhat ambivalent on that, because I'm not convinced that statistics can be racist. But people pointed out to me that the way these statistics are collected might not be free of bias. Also, for example race being a good predictor of creditworthiness might in itself be just an effect of racism (eg. because black people aren't hired at well paying jobs) and using that statistics exacerbates the problem. Those are pretty good points, but I still find it weird that we forbid businesses to use all the information they have available to make business decisions.

View on HN

> Those are pretty good points, but I still find it weird that we forbid businesses to use all the information they have available to make business decisions.

Every rule in society can be formulated as a constraint on businesses. Typically, minimum wage prevents companies from making business decisions that would otherwise be profitable. Banning script money as a payment for wages also prevents them from making these decisions. This stems from the fact that the society view the common good as more important than the business interest of some of its members.

In that sense, making decisions based on data is an action like any other, and is subject to common scrutiny.

The notion that 'common good' is a sufficient reason to restrict individual freedom is debatable. Mainly because it isn't really a definable concept and so 'common good' ends up being a magic phrase to justify almost any restriction small or large.

Your point about minimum wages is a bit misleading. These laws are almost always framed as restrictions on employers but it is worthwhile thinking about them as prohibitions on individuals also. They prohibit individuals from taking jobs that they otherwise would be happy to take. For low-skilled workers (including teenagers) the value of a job isn't just the hourly wage, although $7/hour is better than $0/hour also.

The fact that African Americans, on average, might be significantly less likely to be creditworthy - is not racist per-sey, it's just a fact.

But denying someone credit on the basis of their race is probably racist.

Why? Because having 'black skin' does not make you less credit worthy. It's a correlating factor.

In fact - the 'dangers of stereotypes' are exactly that: making prejudiced decisions based on race etc..

Creditworthiness should be established on the basis of your job, job type, income, education, history of making payments etc..

There's no way on earth it's moral or fair to deny credit because someone is black - in the same way it's ruthlessly unfair to convict someone of a crime because they are black and 'theoretically more likely to have committed crime'.

I'd argue that this is the basis for the systematic racism that exists in the system, and that we have to get over.

And I'm not a pro-PC guy - by far. I'm not even very cool with affirmative action. But we need to treat people as individuals, not stereotypes.

All the factors you listed are also just correlated with paying back a loan. You'd be hard pressed to find something truly causal, human brains are pretty complex.

Having a job vs not having a job + savings is pretty close to 'causal' when it comes to determining the ability to pay back a loan :)

Far, far more so than skin colour :) which is absolutely not causal.

I agree with your point theoretically, but pragmatically ... not so much :)

And a smart ML system would correspondingly put a lot more weight on those variables than on race.

A smart ML system wouldn't put any weight on race. Because we know it's not causal. Also - it's probably illegal, and surely immoral.

>Because we know it's not causal.

How do we know that?

How do we know that skin-colour is not 'causative' to bad credit?

Are you really asking that question?

Not skin color, but race, yes. We already know that there are genes that make certain races predisposed to certain types of behavior. [1] How do we know that no similar genes exist that influence creditworthy behavior?

1. https://en.wikipedia.org/wiki/Monoamine_oxidase_A

FYI - 'race' is a social construct that humans use which basically boils down to skin colour - not much more.

In reality 'African American' or 'Black' should not be though of as a 'race' - rather, as an 'ethnic group' - which is a collection of behaviours, ideals, attitudes etc..

Tell me, what 'race' is Barack Obama?

What 'race' is someone from Sudan wherein people look a mix of 'Middle Eastern' and 'African/Black'?

What 'race' are most Mexicans who have both Aboriginal and European backgrounds?

Genetically, obviously there are some differences beyond skin colour, but it's pretty impossible to pin down 'race' when talking about genes.

The dude doing your credit score is not going a gene test. He's looking at the colour of your skin.

It's insane to think that has anything to do with anything.

If anything 'ethnic group' would be a better correlation, but that's an impossible thing to determine in America.

It's also racist, and I don't mean this in a 'PC' left-wing sense. I mean it in the most basic sense. Which is why it's illegal.

Obama is half white, half black. Sudanese people are a mix of Arabic and African.

Does the existence of gray disprove the notion that black and white are colors?

"Does the existence of gray disprove the notion that black and white are colors?"

Black and White are colours.

People who we commonly describe as 'Black' are not the colour 'Black'.

Tell me - which Black person on planet earth is the 'definitive Black' which equates to '100%' Black.

Barack Obama is 'mostly Black' on one side, 'mostly White' on the other. But his father was from Kenya. Does this make his father 100% Black? Because there are much darker skinned people in West Africa.

Forget colour, it's confusing you. It's a metaphor we use, that is really quite useful colloquially, but not useful beyond that.

There is no such thing as 'perfectly Black' or 'White' genetically. It's just an array of genes. That's it.

There is strong correlation with some genes and some cultures. That's it.

> I'm not convinced that statistics can be racist

This really depends on whether you view racism as a moral failing, in which case the answer is "obviously not" or as a type of bug in a social (or social/technical) system, in which case the answer is "of course".

Statistics collected from a racist society, can obviously reflect the outcomes of racism.

I'm not convinced that statistics can be racist

Statistics don't exist in a vacuum, nor are they a part of nature. Every piece of statistics is the result of a person or a group of people grabbing a subset of some data, applying a collection of mathematical transformations to that data and interpreting the results. Every step of the process, from what data you chose to look and how you chose to collect that data, to what transformations you choose to apply, to how you choose to interpret the numbers that come out of those transformations is a choice a person has to make and as such is very much subject to conscious and subconscious biases.

That's why we have peer review. Are you saying that we should basically discard science because experimental design or statistical analysis can contain errors?

Why do you think black people aren't hired at well paying jobs? Isn't racism also in a way based on stereotypes that are often true? A business owner has no legal obligation to interview every applicant.

That was just an example, I don't really know much about black people in the US.

In a way, it's a somewhat underhanded move - since it's too politically hard to directly help those communities and fix those inputs at the source, we'll create a hidden tax on a bunch of businesses and individuals to do it for us.

On the other hand, if the alternative is no help at all, I can't blame the proponents of those measures.

Reply on news.ycombinator.com