Live data from Hacker News

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

davidrozado.substack.com

511–520 of 697 posts

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

#511

Earlier quoted context omitted.

Not sure i follow, are you assuming i'm in favor of other countries policies that match Israel? To be fair i know little on the subject, and don't claim to. The only reason i said "i can understand .." is that i know it's a hot-button topic. Israel is often portrayed in the west as being overly militaristic towards neighboring areas. The United States itself does similar, depending on who you ask. I judge both quite…

No, I don't mean you personally, I mean the people you might talk to in your circles. Test them by pointing out behavior that they find abhorrent when done by Israel, and see whether they react similarly (by demanding action against them, denying them a right to nationhood etc). If they do: good. If they don't: it's about the Jews, not about the behavior. And from my experience, they almost always don't, and you real…

Fwiw i bet you'd call me anti-Jewish in this context. Not because i believe it more strongly against Israel, but because i'm uneducated on this front and the big countries like Israel make the news more frequently. So if i was to have any opinion at all it's more likely involving Israel than other similar countries.

I suspect that's true for most people.

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

#512

This is something the right leaning groups have always been saying, but it is exciting that the article manages to come up with an empirical way to determine the "belief system" of the AI model. People on the left shouldn't rejoice either. The left and the right wing are mostly the same, and both follow similar approaches to marginalize people who they feel are undesirable, so it'll eventually come back to bite the l…

The other comments show how the demonstrated biases were created. In many well meaning liberals heart, there is a strongly held but rarely publicly discussed belief that they alone belong to the well-meaning, high IQ class. Any other worldview or ideological flavor is always understood by this type as simply incorrect, perhaps caused by failures in morals or intellect. Talk about a buzzkill.

"conservatives understand liberals liberals do not understand conservatives"

>In a study I did with Jesse Graham and Brian Nosek, we tested how well liberals and conservatives could understand each other. We asked more than two thousand American visitors to fill out the Moral Foundations Questionnaire. One-third of the time they were asked to fill it out normally, answering as themselves. One-third of the time they were asked to fill it out as they think a “typical liberal” would respond. One-third of the time they were asked to fill it out as a “typical conservative” would respond. This design allowed us to examine the stereotypes that each side held about the other. More important, it allowed us to assess how accurate they were by comparing people’s expectations about “typical” partisans to the actual responses from partisans on the left and the right)’ Who was best able to pretend to be the other?

>The results were clear and consistent. Moderates and conservatives were most accurate in their predictions, whether they were pretending to be liberals or conservatives. Liberals were the least accurate, especially those who described themselves as “very liberal.” The biggest errors in the whole study came when liberals answered the Care and Fairness questions while pretending to be conservatives.* When faced with questions such as “One of the worst things a person could do is hurt a defenseless animal” or ”Justice is the most important requirement for a society,” liberals assumed that conservatives would disagree. If you have a moral matrix built primarily on intuitions about care and fairness (as equality), and you listen to the Reagan [i.e., conservative] narrative, what else could you think? Reagan seems completely unconcerned about the welfare of drug addicts, poor people, and gay people. He’s more interested in fighting wars and telling people how to run their sex lives.

-"The Righteous Mind: Why Good People Are Divided by Politics and Religion" by Jonathan Haidt p. 334

There's a whole TED talk about it (same author as the above) that goes into the differences in morality: https://www.ted.com/talks/jonathan_haidt_the_moral_roots_of_...

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

#513

Earlier quoted context omitted.

You're responsible for the truthfulness of the statements that you make, especially when the data is right there on the page for you to verify for yourself. "Oh I'm just parroting someone else's distortion as fact" is not a defense. >the picture you're suggesting is itself distorted given that other correlating factors such as social class, are ignored. The picture I'm suggesting is that which the data shows, and you…

But I'm repeating what the actual purveyors of the data said about their own data? They analyzed their data and came to the conclusion you seem to have an issue with- that whites account for 59.1% of violent crimes. In any case, the picture the data suggests is incomplete (because it is missing a crucial data point: social class), therefore the picture you're suggesting is also incomplete.

So if 100 people in group A average a total of 10 crimes, and 5 people in group B average a total of 9 crimes, you think you are being intellectually honest when you say "group A commits more crimes" in the context of comparing the crime levels of those two groups?

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

#514
post #467

Earlier quoted context omitted.

the thing that's always confused me is why would this category of white lumping together latino even be a thing in the first place? Just seems like a really harebrained idea. That association always confuses many hispanics I know during each census. Was there a rationale for this lumping? I never understood it, just seemed to beg for fuzzy/blurred/confusing metrics.

What does "white" mean? Light skin tone, certain physical features? If so, many Latinos are white and that would explain the grouping. Or does it only mean anglosaxon heritage? If so, why?

yah I'm fuzzy on what white is too precisely. I was going to say caucasian but I don't even know what caucasian means exactly. I guess going by today's landscape I think most people would assume European, but that's a pretty broad swath too with lots of variation in it.

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

#515
post #392

Earlier quoted context omitted.

How about the actual article?

I'm sure if you polled society, you would end up with a similar bias. Punching up is generally more acceptable then punching down.

[flagged]

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

#516

Earlier quoted context omitted.

The left's support for Islam directly marginalizes minorities within Islamic countries as well as women and gays fighting for their rights there.

The "left," is a truly nebulous concept if you observe leftist discourse online, where circular firing squads are writ large. I think you are conflating the left's support for minority groups who are demonized and targeted by society for support of the tenets of their respective religions. After 9/11, Muslims were a significant target of hatred and violence. The left, which again is not a monolithic entity, doesn't s…

> .. it supports people who happen to follow that religion. It's based around humanist tendencies, which the right often doesn't understand. You can hate aspects of a religion while also being against any dehumanization and hatred against people who follow said religion.

I quite agree with this statement in the context of what i believe. Which is to say i tend to dislike religions and personally i dislike Islam more than most (based on what i observe), but above all i don't want to target people, Muslim or otherwise, based on their religion.

I think this also supports a sister comment i made where i describe being unsure what my stance is relative to Islam. Observationally it doesn't seem "good" to me, but its people are also target of a lot of hate in this country (USA). It makes it difficult to discuss objective qualities of a religion without it bleeding into racism or bigotry against those who practice said religion.

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

#517
post #408

Earlier quoted context omitted.

It is culture. Actually science on violence and aggression tend to show a bit different picture, and also why such statistics is of such a low value. Good research tend to explore the null hypothesis and alternative explanation for statistical results. A common used data point is that men commit most violent crimes. Using the same data from Sweden one can also say that today Muslim are 20 times more likely per person…

One claim I've seen is that men commit more violent crime because it's more common for a man to be physically stronger than another person, while the rates of non-physial violence are much closer across gender boundaries.

That is one theory I also suspect but I have yet to see a study that look at physically strength during conflicts between people. If it is true then among homosexual relationships the person with less physical strength should be detectable when looking at who is the attacker and who is the victim.

It might also be non-linear. A person who has half the strength of their partner might be significant less likely to be the attacker than if a person has just 10% less physical strength. It would be an interesting read.

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

#518

Earlier quoted context omitted.

This depends on the state, actually. In California political affiliation is a protected status, though how this works out in practice is... variable.

> No, political affiliation is not a protected class in California. A bill that would have made it one failed to pass the state legislature in 2021. [0] Regardless, it shouldn’t be. [0] https://www.shouselaw.com/ca/blog/is-political-affiliation-a...

From the link you posted:

"California employment law forbids political retaliation in the workplace. This includes terminating or retaliating against workers for their political beliefs or activity."

So yes, political affiliation absolutely has strong protections in CA, even if it does not fit a highly technical, but irrelevant, definition which is "protected class".

And the central point is that yes political affiliation has strong protections in CA.

The OP is arguably correct in that they used the words "protected status" which is not a technical term anyway.

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

#519
post #285

Earlier quoted context omitted.

[flagged]

What if they fed the AI crime statistics and it "correctly" identified black people as more violent than other races? What they fed the AI news stories and it "correctly" identified Islam as more violent than other religions?

You mean "correctly" identifying men as more violent than those with less testosterone?

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

#520
post #127

Earlier quoted context omitted.

It's hard to know what "tuned it" means, but they're using an AI model to detect harmful content. So it's very possible that the AI model was always trained on this biased data, but as they made that model more aggressive, it exposed more of it's biases

They're defining harmful content based on political orthodoxy, like every other censorious tinpot dictatorship in history. The objective remains the same too; promote a thought monoculture and propagate the political orthodoxy. The original article makes it very clear that it has nothing to do with preventing harm or promoting equality or any of the other nonsense this is being giftwrapped in.

That's a bit too conspiratorial. The much more mundane reality is that they are defining hateful content based on what is most likely to get breathless, angry, censorious screeds in the media written about how their company is perpetuating (some injustice) if not literally killing people by way of putting the wrong words on the screen.

Culture war outrage drives clicks, it's that simple.

Post reply on HN