Hate Speech Algorithms Racist against Blacks - Politics Forum.org | PoFo

Wandering the information superhighway, he came upon the last refuge of civilization, PoFo, the only forum on the internet ...

Anything from household gadgets to the Large Hadron Collider (note: political science topics belong in the Environment & Science forum).

Moderator: PoFo The Lounge Mods

#15026723
Science man, we have to talk about your racism problem.

A new study shows that leading AI models are 1.5 times more likely to flag tweets written by African Americans as “offensive” compared to other tweets.

Platforms like Facebook, YouTube, and Twitter are banking on developing artificial intelligence technology to help stop the spread of hateful speech on their networks. The idea is that complex algorithms that use natural language processing will flag racist or violent speech faster and better than human beings possibly can. Doing this effectively is more urgent than ever in light of recent mass shootings and violence linked to hate speech online.

But two new studies show that AI trained to identify hate speech may actually end up amplifying racial bias. In one study, researchers found that leading AI models for processing hate speech were one-and-a-half times more likely to flag tweets as offensive or hateful when they were written by African Americans, and 2.2 times more likely to flag tweets written in African American English (which is commonly spoken by black people in the US). Another study found similar widespread evidence of racial bias against black speech in five widely used academic data sets for studying hate speech that totaled around 155,800 Twitter posts.

This is in large part because what is considered offensive depends on social context. Terms that are slurs when used in some settings — like the “n-word” or “queer” — may not be in others. But algorithms — and content moderators who grade the test data that teaches these algorithms how to do their job — don’t usually know the context of the comments they’re reviewing.

Both papers, presented at a recent prestigious annual conference for computational linguistics, show how natural language processing AI — which is often proposed as a tool to objectively identify offensive language — can amplify the same biases that human beings have. They also prove how the test data that feeds these algorithms have baked-in bias from the start.


Vox

A lot of stuff is going on here.

I could see how the reappropriated "N-word" could cause a lot of problems immediately, and perhaps that is the biggest problem here, but it should also give people pause to contemplate whether or not continuing to talk like that online is really appropriate. :hmm:

The article at least seems to talk about it in a way where I feel that it is not just about the N-word.

Sap’s study also tested the bias in these data sets as applied to an open source hate speech detecting tool for developers that’s run by Jigsaw, a subsidiary of Alphabet (Google’s parent company). The open source tool, called PerspectiveAPI, is used by news organizations such as the New York Times to help moderate comments online. It’s the publicly available version of an underlying technology that’s used throughout Google for its own products as well.

Researchers found that data run through PerspectiveAPI showed a significant bias against African American speech, labeling those tweets as toxic more often.


The labeling of something as 'toxic' seems to indicate that they aren't just looking for a single word. At least, in my perspective, they would simply talk abotu the presence of obscenity if that was the case, but IDK, I could be over-thinking it and speaking without evidence.

The researchers and others have advocated for giving moderators more social context about the people writing tweets, but that can prove tricky. When Facebook’s content moderation guidelines are already under scrutiny, would giving moderators more context open the door for more criticism, particularly from conservatives? And when content moderators at Facebook and other companies are reportedly working under grueling conditions and are pressured to rate content as offensive or not, making more nuanced decisions could make a difficult job even harder.


... More social context? Fancy that.

I have found that when people are advocating to give very special attention to something, they are usually looking for a way to justify it.

Nuance, after all, requires space and room.
#15027111
Humorous. I had posted a year or two ago about AI HR bots being racist. They ended up tossing the resumes of black candidates far faster. I'm guessing the issue with hate speech is that they are applying standards uniformly, and not taking race into account and lightening the censure for minorities. After all, it's okay to be a misogynist if you're black or a Muslim. It's only wrong if you are white or a Christian in modern Marxist theory.
#15027127
AI cannot read tweets in context as it can only identify offensive terms, which results in false positives (FP). AI cannot guess the author's race, too. For instance, the N-word is offensive to African Americans, if it is used by non-blacks. But the term is acceptable within the African American community. Human moderators are necessary to get around the problems.

Image

Results Figure 2 (left) shows that while both
models achieve high accuracy, the false positive
rates (FPR) differ across groups for several toxicity
labels. The DWMW17 classifier predicts almost
50% of non-offensive AAE tweets as being offensive,
and FDCL18 classifier shows higher FPR for
We further quantify this potential discrimination in
our two reference Twitter corpora. Figure 2
(middle and right) shows that the proportions of
tweets classified as toxic also differ by group in
these corpora. Specifically, in DEMOGRAPHIC16,
AAE tweets are more than twice as likely to be
labelled as “offensive” or “abusive” (by classifiers
trained on DWMW17 and FDCL18, respectively).
We show similar effects on USERLEVELRACE18,
where tweets by African American authors are 1.5
times more likely to be labelled “offensive”. Our
findings corroborate the existence of racial bias in
the toxic language datasets and confirm that models
propagate this bias when trained on them.8
https://homes.cs.washington.edu/~msap/p ... 19risk.pdf
#15027143
ThirdTerm wrote:AI cannot read tweets in context as it can only identify offensive terms, which results in false positives (FP).

Yes, but AI picks up those terms too.

Princeton researchers discover why AI become racist and sexist

Ever since Microsoft's chatbot Tay started spouting racist commentary after 24 hours of interacting with humans on Twitter, it has been obvious that our AI creations can fall prey to human prejudice.


The more you analyze, the more you will find differences. For example, my ice maker isn't working and I wanted to get to the store early this morning to get ice before the crowds came. The customers at the store near my house are very typically mixed race. Yet, when I went very early in the morning, almost all of the customers were white. I've stated that I notice these things on mass transit too--such as the racial admixture of the BART train passengers at the Oakland Coliseum station versus the airport tram. The BART train is thoroughly race mixed, but the airport tram is mostly white. People are very uncomfortable with talking about things like that, but a computer algorithm doesn't care at all. It's not self-conscious or socially conscious. It's just making inferences.

As you noted, an n-word that is socially acceptable among blacks gets flagged by AI; yet, it isn't considered offensive if used by blacks, or even when used by whites they know socially. That reveals its own set of biases--in asserting equality among people, censors insist that rules do not apply equally to all people. In other words, in purporting to oppose inequality, they are in fact defending inequality.

Israel won't comply because one of their explicit[…]

Commercial foreclosures increase 97% from last ye[…]

People tend to forget that the French now have a […]

It is easy to tell the tunnel was made of pre fab[…]