Math is Racist

Duh, it's been engineered by misogynistic CIS white males to be incomprehensible to minorities.

#boycottmath #firetheprofessors #thisishowwefightbigotry #sjw2016 #timetogetridofsocialconstructsandarbitrarybarriers #idtellyoutogoeducateyourselfbutyouliterallydonthaveabrain #BLM #breaktheglassceiling #lgbtqpfxyzRULEz

>posting an article about math in mainstream media
>/pol/

Yeah, no.

The "social sciences" are coming for our real sciences all the same.

yes it is

While the title of the article is totally bogus and inflammatory, the article does have a point: statistical methods can very easily make deductions that are totally valid, yet still against your rights. For example, if your siblings and parents all have a criminal record involving theft, then I can make a perfectly valid deduction that you are a more-likely-than-usual suspect. But to hold this against you in a criminal investigation would be a violation of your rights, despite the fact that the deduction is a valid one. Decision making based on statistical models appears to generally be not nearly careful enough to avoid this.

>But to hold this against you in a criminal investigation would be a violation of your rights
Yes but no one is using or suggesting we use predictive algorithms in that way.

This is a clickbait title for an article. Im 100% sure anyone in this thread actually read it. Go post this on /pol/. Maybe you will get rage comments galore.

...

article basically is just saying that you can't just look at numbers to evaluate things without utilizing context

guns dont shoot people.

You can if you have enough numbers to look at

I can sense samefag

You are wrong. The choice of numbers gathered from statistics implies a creation of a correlation between facts based on personal interpretation.
Pic related.

well the chief criticism in the article is over reliance on zip codes

if you live in a zip code with people of lower credit score, your credit score will automatically be lower. so people who are turning 18 and are in these poorer areas automatically have a lower credit score, sometimes significantly. with a lower credit score, it is more difficult to find a job, take out a loan for a car/get a business, etc. this enhances the risk for poverty and continuing the cycle

where race comes in is that these "red flagged" zip codes are more likely to be minority-heavy than others for a variety of historical reasons

Exactly why you can't just line up two graphs () and look for a correlation.

If you want to make good predictions, you need many many different sets of data.

yeah and I'm not frankly sure there really is enough collectable data to get a fully good credit assessment of an individual

on one end there are obvious issues with factoring zip code into things like credit score but on the other hand you don't want to just outright lend at low interest rates to everyone

Correlation does not equal causation

You can take dozens of graphs with the same shape and plot them together. It still means absolutely nothing until you use personal interpretation to derive a cause-effect relation between two facts. There are a lot of graphs that have the same predictable curve, specially in economics, and often lots of them follow the same rate of change without necessary correlation.
The point is that statistics mean nothing until you create an interpretation based on them, and often this interpretation has emmotional and ideological bias. It is extremelly easy to create false correlations between events that follow the same variation curve, because variation curves are usually predictable and not at all random, since they follow common patterns.

...

You don't need to know the cause and effect relationships in the data to know if someone is unlikely to be able to pay off a loan.

But can you define causation relations in any case using only statistical data? A causation relation arrives only from personal interpretation of facts when it comes to data. This is the meaning of the picture.

Of course you need to establish a causation relation if you want to predict a behavior. This is the logical basis of it. In order to make sense of the data, you need to take those numbers and transform them into a function to be able to predict the behavior of it through time.
If there is no causation relation between, say, living in a certain area and being able to pay loans, then there is no logical meaning to deny a loan to a resident of this area. You can only do that if you establish a relation and predict the behavior of that event based on the function you established.

I don't even think it's legal to consider zip code in credit score for the exact reasons she states so I'm not even sure where she's coming from

en.wikipedia.org/wiki/Community_Reinvestment_Act

>need to establish a causation relation if you want to predict a behavior
You literally don't

They do establish general causations in these instances - lower income, less steady employment etc. Problem is tangential to this - the fact that race also often correlates with it. The argument being made in the article is that it's sort of a self-repeating thing in the end - the low credit scores keep people down because it makes it more difficult to get a job and so on

It does suck for some people, but banks aren't charities.
They lend money to make a profit, not out of the goodness of their hearts in the hope that you are one of the 5% that breaks their data prediction.

A causation relation is a function f(a,b,...,n) of n variables.
Give me one way of calculating the value of a variable, say, "b" with given values for the other n-1 variables without using a function of the form f.

Oh yeah I understand that

That's why I think it would be good if the communities either formed some sort of collective financing organizations (like the co-ops we have here in Germany) that were self-supporting or had some microfinance charity arrangement

I'm saying you don't need to know why the variables are what they are. You just need the predictive relationship between them.

I don't need to know why you live in a ghetto, or why you didn't graduate high-school, just that the predictive relationship says given those variables, you are 95% unlikely to pay off a $5,000 loan.

...

>I think it would be good if the communities either formed some sort of collective financing organizations
we have those. they're called credit unions

They don't have them in the ghettos as much, but I read recently they have grown hugely in popularity recently

The algorithms that factor in flagged zip codes as a negative impact on one's ability to pay back loans isn't just grabbed out of thin air. Sure, it's not fair to the hard-working individual cursed with the zip code he or she lives in, but the main objective of these results is to determine a probability of the borrower's ability to repay. If people from area X are generally bad at repaying loans, then anyone from area X is more likely to also be bad at repaying loans. It's not the fault of anyone--and every lender has the opportunity to give a bit more of an ear to his or her prospective borrower. That's where the idea of these gross "generalizations" are made. The math didn't do anything, it's the people who determine it to be more than it is that end up perpetuating the "cycle of inequality" or whatever bullshit they want to call it. My bet is O'neil had her thesis written by her advisor.

>X is a problem for Y
>alot of Z are Y
You might think they would write it as:
>poor people negatively impacted by zip code analysis
But why would you do that if you can combine clickbaiting and racebating to create a wonderful headline like:
>math is racist

Probably, why read click bait though?

yes this is media 101, always phrase the title of an article in such a way that it gets people to click on it, even if they dont read it.

Without math there would be no monetary system and no capitalism, which are responsible for all problems in the world. Explain yourselves mathfags.

>article has a provocative title
>you get provoked by it
Well done OP, you let them win

i dont get it

don't put your idiotic tabloids here without an archive link

Praise kek!

>You literally don't

Better start eating then.

> (OP)
>
>
> (You)
>I can sense samefag

Your senses suck

No. Lets just give loans to people that can pay them back in full and on time. We have limited resources on the planet, lets allocate them efficiently.

you've made it obvious you don't understand statistics.

you can predict with no casual relations at all.

Next year people will complain how biology is racist because we aren't all white

these models are adopted because they work!

they're literally the result of optimisation!

statistics is really just an application of optimisation.

how can this woman argue against their use when she must understand this?

on moral grounds?
well then the choice is between whether you morally prioritise having a system that works efficiently and ditributes police resources in a way that catches and prevents the most crime or whether you have a system that is "fair" because it ignores information that would help it work more effectively.

how is this a dilemma??
why would you prioritise not making good use of data?!

this is extremely dangerous thinking.
how unsurprising that such emotional thinking came from a woman.
It's like even when they possess a strong mind capable of statistics they still fall to soft-brained, emotional appeals.

This is called data-snooping (among other things) and is easily avoided testing your hypothesis or model against a different set of data from the one that you did exploratory data analysis with in order to come up with your hypothesis.

What are you hoping to point out with these examples?

>these models are adopted because they work!
Fuck no they aren't.
They're adopted because they look good on paper, and the people pushing them are good enough at marketing to find people to con.

>why would you prioritise not making good use of data?!
Why are you assuming that these models "make good use of the data"? Throwing some shit together that makes pretty looking graphs is vastly cheaper than careful statistics.