Are Algorithms Racist?

H

HardOCP News

[H] News
Joined
Dec 31, 1969
Messages
0
This entire article talks about the discriminatory problems associated with algorithms and what can be done about it. Racist algorithms? Is the author onto something or just on something?

Another solution, says Venkatasubramanian, is to screen algorithms more rigorously, testing them on subsets of data to see if they produce the same high-quality results for different populations of people. And Crawford says it might be worth training computer scientists differently, too, in order to raise their awareness of the pitfalls of machine learning in regards to race, gender, bias and discrimination.
 
Algorithms profile by design and if one of the available inputs is race than it can indeed produce results that are biased that way. However, if Gender/Race X/Y is Z% more likely to do or be affected by something then how exactly is it wrong to use that information as long as the statistics are constantly updated?

It's odd that we openly accept this bias is medical diagnosis, but shun it for other purposes. The data doesn't care about your religious or political leanings and is solely based on what has been done to predict what may happen if the trends continue.

I think the main issue is that it cuts through the Political Correctness BS and highlights the truth underneath the smoke being blown up our collective a****s.
 
Data should be processed as strictly as possible.
Search the data, these are the results.
If they are racist in nature, then that is indicative of what is being measured, not a flaw in the algorithm.
Arbitrary, human biased decisions should be avoided when interpreting data. Processing data should be as boolean as possible. If the data is showing patterns in race, then, that is indicative of what is actually happening. Fudging these numbers to fit ultimately some PC overview is stupid and will result in inefficiency.

If the data shows race specific results, USE THEM, do not hide them.
 
Taldren said:
Algorithms profile by design and if one of the available inputs is race than it can indeed produce results that are biased that way. However, if Gender/Race X/Y is Z% more likely to do or be affected by something then how exactly is it wrong to use that information as long as the statistics are constantly updated?
Click to expand...

Is it wrong to not get hired because racist policies of the past had an effect on your score regardless of what your race is? (See: redlining)
 
Kalabalana said:
If the data shows race specific results, USE THEM, do not hide them.
Click to expand...
I've heard it said that it's only okay to use if it pertains to the (dwindling) white majority, if about any other group, it's (institutionalized?) racism.

evilsofa said:
Is it wrong to not get hired because racist policies of the past had an effect on your score regardless of what your race is?
Click to expand...
I've never seen a worse crime in the entire world than a group of similar people favoring their own.

#buyblackeconomics
 
Maybe our definitions for what is racist and discriminatory needs to be changed...
 
xJ321x said:
I've heard it said that it's only okay to use if it pertains to the (dwindling) white majority, if about any other group, it's (institutionalized?) racism.


I've never seen a worse crime in the entire world than a group of similar people favoring their own.

#buyblackeconomics
Click to expand...

Truth of the matter is, ALL groups are biased by "nature" -- I am closer to my g/f than some chick on the street, chick on the street is closer to her family than an Inuit, Inuit is closer to his friends than a guy in Russia, etc ... nothing wrong with that, by and large. A world where everyone is eating McDonald's all the time would be nightmarish ;P

It's when groups of people claim that their overall inherent bias is superior to the overall inherent biases of others, that miscommunication and racism appear ...
 
evilsofa said:
Is it wrong to not get hired because racist policies of the past had an effect on your score regardless of what your race is? (See: redlining)
Click to expand...

If this were the case, only Asian's would be getting hired. Plus is may in fact not have anything to do with race or gender and still seem like it.

Lets say School A doesn't prepare students well for the real world by acting more as a social experiment rather than a institute of higher learning. The algorithm is going to degrade the score of all those that attend School A and rightfully so. Now if School A has a skewed demographic then on the output it may seem to show racial and gender bias when it never even considered those as an input.

To me this seems more like an attempt to escape the consequences of policy rather than anything else.
 
tungt88 said:
ALL groups are biased by "nature"

It's when groups of people claim that their overall inherent bias is superior [even in their own areas] to the overall inherent biases of others, that miscommunication and racism appear ...
Click to expand...

What I rationalize from all of this is to blame the winning group for discrimination until they're no longer the winning group. Then the new winning group will do the very same thing and favor their own group. Why? Because everyone is 'naturally biased'.
 
Statistics are often also biased and racist despite being just plain statistics. The politically correct will blow a vent on that fact.
 
Data is not racist or biased; those are human constructs. The way that data is collected and presented will be tainted by the person handling it or by the observer, revealing their own personal biases in its interpretation.
 
Steve said:
This entire article talks about the discriminatory problems associated with algorithms and what can be done about it. Racist algorithms? Is the author onto something or just on something?

Another solution, says Venkatasubramanian, is to screen algorithms more rigorously, testing them on subsets of data to see if they produce the same high-quality results for different populations of people. And Crawford says it might be worth training computer scientists differently, too, in order to raise their awareness of the pitfalls of machine learning in regards to race, gender, bias and discrimination.
Click to expand...

An algorithm is just a process. If the process is racist then an algorithm to perform that process will be racist, the same is true in the opposite.
 
What the hell?!?!?!?!?! Next the you know, I will not be able to call the color of my car black. (Yes, it is black and no, I could not care less what you think.)
 
evilsofa said:
Is it wrong to not get hired because racist policies of the past had an effect on your score regardless of what your race is? (See: redlining)
Click to expand...

It's wrong not to be able to hire the most qualified people for your job regardless of why there is a difference in qualifications.

A flawed foundation will always poorly support the structure that rests on it. You can only fix things properly from the beginning no matter how badly you want to fix them in the immediate.

If you want to fix racial inequality when it comes to opportunity, then you have to do it from the start, you can't bastardize an entire system for immediate remedies.

Now that is what we have done here in the US. We pushed for immediate reforms and used quotas and every other kind of shoe-horned solution to fix inequality in the present and instead of being fair we were just unfair to a different group of people.

The persistent application of unfairness perpetuates racial problems, it doesn't eliminate them. You can't treat people fairly by treating them differently.
 
The best solution to this problem is to try and create a robust economy where there are more "good" jobs then there are applicants (rather than the reverse situation we have now) ... if there are enough jobs so that everyone who wants a job has one, then many of the racism questions go away as the workers can find the jobs and companies and responsibilities that are best suited to them (regardless of their race, sex, experience, age, or education)
 
The best solution for whom?

Not for the companies, they want a workforce that's in competition with itself so that they can pick from the best and not have to pay for lesser talents.

In the same manner that companies compete in order to reach a position of dominance, and therefore no competition, they want a workforce that follows suite. This is how companies ensure that their workers mirror the company's "core values"

See what I did there? :D
 
lcpiper said:
The best solution for whom?

Not for the companies, they want a workforce that's in competition with itself so that they can pick from the best and not have to pay for lesser talents.

In the same manner that companies compete in order to reach a position of dominance, and therefore no competition, they want a workforce that follows suite. This is how companies ensure that their workers mirror the company's "core values"

See what I did there? :D
Click to expand...

I'm still old school and like to look for the win-win solution :cool: ... in the 90's we had a seller's market where there were more jobs than workers (good for worker wages and worker mobility but bad for companies) ... in the 2000's we flipped to a buyer's market where there were more workers than jobs (good for company profits but bad for workers) ... I can hope for a mythical (but likely impossible) world where everybody has the perfect job and workers and companies are both reasonably balanced and happy
 
I agree, but I think what it needs is for proven companies to help smaller companies so that new startups are encouraged. We have that to a degree today, I just don't think enough people are willing to seize the opportunities that are available to them. Being the entrepreneur is not a common thing.

And with that, to any of you who have done so and started your won businesses, my hats off to you :cool:
 
lcpiper said:
It's wrong not to be able to hire the most qualified people for your job regardless of why there is a difference in qualifications.

A flawed foundation will always poorly support the structure that rests on it. You can only fix things properly from the beginning no matter how badly you want to fix them in the immediate.

If you want to fix racial inequality when it comes to opportunity, then you have to do it from the start, you can't bastardize an entire system for immediate remedies.

Now that is what we have done here in the US. We pushed for immediate reforms and used quotas and every other kind of shoe-horned solution to fix inequality in the present and instead of being fair we were just unfair to a different group of people.

The persistent application of unfairness perpetuates racial problems, it doesn't eliminate them. You can't treat people fairly by treating them differently.
Click to expand...

Well said. It not only perpetuates, it exacerbates the issues. The equivalent of multiple Band-Aids and painkillers don't "fix" the core issues ...

I have my own business. The vast majority of my income is derived from partnering with a much larger business. They help us, and we help them. Works out well for the both of us, and both businesses have been making a lot more money than they were before we partnered together :)
 
Mainly you can make anything racist to a certain extent. Algorithms are really just instructions. I mean sure you could right a class that has racial overtones as far as hair color, eye color, etc., but I'm not sure that makes it racist.
 
Everything is racist.
Everything is sexist.
You are always oppressing minorities and women at all times. Even when you're coding.

If you are a male Caucasian you are committing these acts at all times, and we need to point them out and recognize them. I propose putting all white males that are oppressing others in camps designated specifically for them. This will help create a "safe space" for us where there aren't any facts to be heard or get in the way.
 
Cyraxx said:
I propose putting all white males that are oppressing others in camps designated specifically for them. This will help create a "safe space" for us where there aren't any facts to be heard or get in the way.
Click to expand...

That wasn't in the Trump press statement. You made that up.
 
160111_max.jpg
 
Kalabalana said:
Data should be processed as strictly as possible.
Search the data, these are the results.
If they are racist in nature, then that is indicative of what is being measured, not a flaw in the algorithm.
Arbitrary, human biased decisions should be avoided when interpreting data. Processing data should be as boolean as possible. If the data is showing patterns in race, then, that is indicative of what is actually happening. Fudging these numbers to fit ultimately some PC overview is stupid and will result in inefficiency.

If the data shows race specific results, USE THEM, do not hide them.
Click to expand...

But that is what PC is about. Falsify or sweep the truth under a rug, to make things look better.
 
The title is click bait, the results are based upon certain criteria. It'd be like picking 3 guys out of 10 and the results happen to pick 3 of the same race. The process isn't racist, it's the interpretation that is.
 
You must log in or register to reply here.
Back
Top