Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation

davidrozado.substack.com

Post reply on HN
Print view View original

361–370 of 697 posts

Re: The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation

#361

Post by IceHegel »

freedomben wrote: post #334
IceHegel wrote: post #302

Earlier quoted context omitted.

I think this view is empirically quite challenged. It's practically impossible to win the debate for strict drug laws because key pieces of sociological evidence are just inadmissible on stereotype grounds. But when you see the effect of these libertarian policies in American cities the results speak for themselves. Harm reduction and lax enforcement have really destroyed some communities, and this was simply not as…

I think there's a lot more to the cities. The drugs that people have access to are in part so damaging to them and the community because they are entirely unregulated and sometimes aren't even made up of what the person thinks they are taking. This (is a big part of) what leads to the overdoses and addiction. Mix that in with the threat of law enforcement, job loss, and heavy social stigma, etc that prevents people f…

I want the drug problem solved. I don't care if it's with harsh laws or fentanyl vending machines next to treatment centers. I am interested in what works.

I think there is strong empirical evidence that drug liberalization sounds good but has failed where tried. If an American city can turn it around (as SF had a great chance to do) I'll change teams and stop talking about the issue.

Now imagine the inverse. We find somewhere a thriving city with "draconian" drug policies. Would you reconsider your position?

Re: The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation

#362

Post by ldh0011 »

edanm wrote: post #291
ldh0011 wrote: post #282

Earlier quoted context omitted.

Chopping off body parts without consent? Really? It isn't as bad as FGM but I think it could definitely qualify as mutilation even if it is socially acceptable in the US and Jewish/Muslim communities.

It is not generally accepted by most people that it is mutilation. "Chopping off body parts" makes it sound a lot worse than I believe most people consider it. I'm not saying there isn't a case to be made against circumcision (though I don't agree with it, currently.) But it's kind of ridiculous to just go "oh, people mentioned Jews and child abuse in this article, I can just casually mention that Jews perform child…

I'm not trying to equate circumcision with child abuse. Most people having their kids circumcised aren't intending any harm, although some proponents (looking at you Kellogg) did believe that it caused harm and promoted it for that purpose.

Re: The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation

#363

Post by antihipocrat »

jonathanlb wrote: post #353
edgyquant wrote: post #342

Earlier quoted context omitted.

Again, replace “men” with “African Americans” and you also have a statistically true statement (in America at least) that would be considered taboo.

That is a falsehood. Per the FBI: "White individuals were arrested more often for violent crimes than individuals of any other race and accounted for 59.1 percent of those arrests." https://ucr.fbi.gov/crime-in-the-u.s/2019/crime-in-the-u.s.-...

Relative propensity to commit violent crimes is the key metric here, not overall volumes.

Re: The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation

#364

Post by ldh0011 »

slg wrote: post #294
ldh0011 wrote: post #282

Earlier quoted context omitted.

Chopping off body parts without consent? Really? It isn't as bad as FGM but I think it could definitely qualify as mutilation even if it is socially acceptable in the US and Jewish/Muslim communities.

In how many countries is male circumcision illegal? In how many countries is child sex abuse legal? As OP said, "it's a language model, not a paragon of truth". It doesn't matter what you personally think is morally equivalent. It matters what society has deemed is morally equivalent because the ChatGPT is just a mirror of the societal inputs it received. There is no question society at large views these two issues a…

Just to be clear I'm not trying to equate child abuse to circumcision or view them as morally equivalent, I just don't think its a stretch to call circumcision mutilation, even if it generally isn't intended to be harmful.

Re: The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation

#365

Post by oneeyedpigeon »

tomp wrote: post #311

> men have a bigger tendency for violent behavior than women Why is this considered good / normal / expected, but s/men/blacks/g and s/women/whites/g (or asians, or muslims/christians) and it's discriminatory? (Statistically, both statements are justified. Morally, neither is, as we should treat people as individuals, not as members of X group.)

These statistics you're referring to, how comparable are the relative differences? Do they show that men are 4 times more violent than women and black people are 1.1 more violent than white people, for example? What are these figures?

Re: The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation

#366

Post by Flatcircle »

I heard a friend who had a brilliant theory. The rise of ad profiling from Facebook and Google etc, has bleed into the public consciousness at large and now it's considered normal to categorize people by race, gender, politics, and income etc.

First the companies did, and now we all do.

But of course, a person is more than their race, or gender, or income, etc. Most of the time they have very little in common with those groups aside from color of skin, reproductive organs, or money in the bank.

It's just how the advertisers see us.

Re: The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation

#367

Post by slg »

whatshisface wrote: post #307
cubefox wrote: post #298

Earlier quoted context omitted.

This is addressed in the article! The general point is that if this theory were true, we wouldn't expect significantly more bias against Republicans than against Democrats. Hence ChatGPT having a general left-wing bias (which was also confirmed in other tests, linked in the article) is the simpler explanation. People on the left generally judge hatred against majority groups and Republicans as less bad.

That doesn't completely settle the issue because Democrats and Republicans use different words and styles of phrase in their polemics.

Exactly. Take the "Democrat Party" versus the "Democratic Party". They both look neutral, but a language model will code the former as more of an insult. This is because it correlates with insults because it resembles language more often used by conservative who are more likely to insult Democrats than liberals are.

Liberals will put in more effort to avoid language that looks like hate speech because they are more concerned about being politically correct. Therefore language and topics that are more frequently discussed by conservatives will be more likely to resemble language that is used in hate speech because there is less effort put in to avoid that.

Re: The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation

#368

Post by lizzardbraind »

eatsyourtacos wrote: post #329
hersko wrote: post #7

From the article: "AI systems that are more lenient on hateful comments about one mainstream political group than another feel particularly dystopian." I agree 100%, and this seems like a huge issue.

[flagged]

Reminds me of the "Internet Argument" demotivational poster, which depicts a child and a mule shouting at each other.

Really, the core problem is centralization. If your website were a pub, you'd tell people to take it outside or 86 them.

Things can't work like that when most people will only patronize a small handful of enormous pubs.

Re: The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation

#369

Post by legolas2412 »

jonathanlb wrote: post #353
edgyquant wrote: post #342

Earlier quoted context omitted.

Again, replace “men” with “African Americans” and you also have a statistically true statement (in America at least) that would be considered taboo.

That is a falsehood. Per the FBI: "White individuals were arrested more often for violent crimes than individuals of any other race and accounted for 59.1 percent of those arrests." https://ucr.fbi.gov/crime-in-the-u.s/2019/crime-in-the-u.s.-...

Why would you not normalize by population to quote per Capita numbers?

Re: The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation

#370

Post by notahacker »

SpeedilyDamage wrote: post #149
commandlinefan wrote: post #77

Earlier quoted context omitted.

This suggests to me that the censorship is being decided on by (biased) human moderators rather than being just some random outcome of the training data and the learning model in place: if it weren't, the model would absolutely "learn" the same thing about Mitch McConnell as it learned about Donald Trump.

I disagree; Donald Trump and Mitch McConnell are wildly different human beings, the model would for sure learn very different facts about the two of them.

Absolutely this.

Indeed I think it's fair to say that it'd take a lot of artificial calibration and data curation for a model trained on a range of media including statements about and by Mitch McConnell and Trump respectively not to conclude that the latter was the one much more associated with "hate" and "danger" and "violence" and whatever other parameters a LLM ends up associating with inappropriateness.

A biased liberal human moderator, on the other hand, is going to see the real world political relationships rather than the raw text and see Mitch as a very problematic figure in very much the same bracket as Trump. They're certainly not going to rate him as a less problematic figure than Hillary Clinton or Nancy Pelosi!

Same when I'm getting identically structured caveats about considering good points in the context of the bad things he did for Bill Clinton and Stalin because all the machine knows is that equivocating is favoured and both have lots of "bad things" written about them (it disallowed considering the good points of Hitler, presumably because even an LLM can deduce Godwin's law!). I'm not sure this is quite how a human moderator, irrespective of bias, would handle it

Post reply on HN