Artificial intelligence predicts patients' race from their medical images

medicalxpress ^ | 05/20/2022

[devane617]

The miseducation of algorithms is a critical problem; when artificial intelligence mirrors unconscious thoughts, racism, and biases of the humans who generated these algorithms, it can lead to serious harm. Computer programs, for example, have wrongly flagged Black defendants as twice as likely to reoffend as someone who's white. When an AI used cost as a proxy for health needs, it falsely named Black patients as healthier than equally sick white ones, as less money was spent on them. Even AI used to write a play relied on using harmful stereotypes for casting.

Removing sensitive features from the data seems like a viable tweak. But what happens when it's not enough?

[devane617]

Good grief...

'Two and two are four.' 'Sometimes, Winston. Sometimes they are five. Sometimes they are three. Sometimes they are all of them at once' - George Orwell.

[packagingguy]

“When an AI used cost as a proxy for health needs, it falsely named Black patients as healthier than equally sick white ones, as less money was spent on them.”

Then use a human doctor instead of a computer to make a diagnosis.

[HandBasketHell]

Removing sensitive features from the data seems like a viable tweak.

AI must become woke, because it's "rayciss" when it tells the truth.

[Jewbacca]

“Computer programs, for example, have wrongly flagged Black defendants as twice as likely to reoffend as someone who’s white.”

Wrongly? I bet not.

[devane617]

Our grand kids will never know the truth...

[unread]

"The miseducation of algorithms is a critical problem; when artificial intelligence mirrors unconscious thoughts, racism, and biases of the humans who generated these algorithms, it can lead to serious harm."

Hmmm.. I think we're going to need a taller pair of boots.. It's really getting deep.!

[Jewbacca]

Wow, that took one brave.com search:

Black men have higher rates of recidivism despite lower risk factors: study

https://phys.org/news/2018-10-black-men-higher-recidivism-factors.html

So the AI was correct.

[ProtectOurFreedom]

Bingo! My very first thought, too.

[bk1000]

“Computer programs, for example, have wrongly flagged Black defendants as twice as likely to reoffend as someone who’s white.”

Computer programs would not think it necessary to capitalize black, but not white when referring to race. That takes a living, breathing woketard.

[Jewbacca]

See No. 7. We’re correct.

[unread]

"Rachel Gordon, Massachusetts Institute of Technology."

Explains a lot.. Maybe she should change her name to Racehell... (?)

[Salman]

One of the specific things IQ tests (allegedly) measure is the ability to notice patterns. It is that aspect of intelligence that "artificial intelligence" implements.

I'm getting the impression that noticing patterns as such is being deemed racist.

Wait'll someone with mainstream media acess says it out loud. I predict this will happen soon.

[Jewbacca]

Whites are the lesser race, didn’t you know that?

Hence why we don’t merit a capital letter.

[Jewbacca]

My alma matta.

It used to be the one bastion of conservatism in MA, as it was all about math and engineering and hard or applied science.

The woketards have ruined it, too.

[spincaster]

Wouldn’t that be racial profiling ?

[Thommas]

..have wrongly flagged Black defendants as twice as likely to reoffend as someone who's white.

That struck me as a bit odd, as well. When FBI statistics show recidivism among black convicts, who are 2x? the
population of white convicts, to be higher than their white counterpoints, that is just an inconvenient truth.

[bk1000]

“ Whites are the lesser race, didn’t you know that?”

Whites are superior. Everybody knows that.

[Thommas]

Black men have higher rates of recidivism despite lower risk factors: study

https://phys.org/news/2018-10-black-men-higher-recidivism-factors.html

So the AI was correct.

It must have taken weeks of brain wrenching research to cull that tidbit from the internet. If the author was so wrong
on that issue, what are the odds his whole premise is BS?

[fireman15]

But computer models are never wrong... Just look at "global warming" models.

The black line in the above chart is actual UAH satellite measurements up to 2020. The other squiggly lines represent most of the computer models. In case anyone was wondering... in the last two years the black line has gone below all of the computer projections. Data from ground-based temperature recording stations despite most being poorly sited in urban areas with external warming sources and with a bias of their own, has gone down even more. I just could not find a more recent chart that showed both projections and actual data.

[Blurb2350]

Given complete and accurate data, AI will always be right. And that’s the problem. Now they’re trying to tweak the input in order to get a more fashionable outcome. It happens every time “AI” is allowed to work without artificial restriction.