The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation • Hacker News Board

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

#671

Post by antifa » Sat, Feb 04, 2023, 1:56 PM UTC

ericmcer wrote: post #304

npteljes wrote: post #135

Earlier quoted context omitted.

>I think most people would agree that a lot of the content on the internet is left-leaning. This is an empty statement. The internet is vast, and it's filled to the brim with content, from all sides. For a demonstration, check the linked website, and click the Bias categories. Huge amount of publications, in every category. https://mediabiasfactcheck.com

Do you really think counting the number of websites of a certain bias and comparing the two counts is an accurate measure? How many of those websites do you think it takes to equal the readership of a single Reddit or Fox News? We know most young people are left leaning. We know most content on the internet is created by young people. It isn't a huge leap that the internet is mostly left leaning. If you don't believe…

Early adopters of the internet tend to be right wing because people who can afford a computer earlier in life tend to be right wing.

The international right wing has not finished being an early adopter in their respective countries, which means the international left wing will follow shortly after.

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

#672

Post by throwaway894345 » Sat, Feb 04, 2023, 3:00 PM UTC

SpeedilyDamage wrote: post #664

throwaway894345 wrote: post #659

Earlier quoted context omitted.

They are both on wikipedia; I read them yesterday. On mobile now, but you can google them easily. They don’t have anything to do with conservatives views being suppressed because we weren’t talking about that. The original claim was that people who get “called out” have reprehensible views—I gave examples of people who were fired or harassed or etc for completely innocuous speech.

…what do you think the title of this submission is?

Oh dear, you think every comment on HN is a direct reply to the title? No, comments are arranged in a tree structure. For example, the comment you're reading right now is a reply to your previous comment, not to the title. It's not just HN that structures its comments this way--it's a very popular schema, so this bit of knowledge should be really helpful for you on your Internet adventures. Godspeed!

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

#673

Post by throwaway894345 » Sat, Feb 04, 2023, 3:09 PM UTC

SpeedilyDamage wrote: post #665

throwaway894345 wrote: post #661

Earlier quoted context omitted.

Agreed. Quite a lot of people think deviating from the party line is reprehensible, but these things are all squarely within the Overton Window irrespective of mine or your opinion. I’m curious about why you find Fang, Shor, etc reprehensible though.

I’ve said multiple times I cannot find the controversy you keep claiming happened to those two people, which to me speaks more loudly than the actual controversy they supposedly had.

I pointed you multiple times to their Wikipedia pages. Here are some direct links to the specific relevant sentences (I didn't post them before because I was busy and on mobile and I didn't realize how burdensome you would find it to find and skim their wikipedia pages). I'll quote the relevant sections as well in case you have a hard time working those links.

https://en.wikipedia.org/wiki/Lee_Fang#:~:text=In%20June%202...

> In June 2020, Fang was accused of racism by Akela Lacy, a colleague at The Intercept. This occurred after Fang shared a Martin Luther King Jr. quote about remaining non-violent and tweeted out an interview in which a black man at a George Floyd protest expressed concern about black-on-black crime. Fang's tweets set off a "firestorm" on Twitter and he issued a lengthy apology

https://en.wikipedia.org/wiki/David_Shor#:~:text=On%20May%20....

> On May 28, 2020, Shor tweeted a summary of an academic study by Omar Wasow, a black political scientist at Princeton University, that argued riots following Martin Luther King Jr.'s assassination likely tipped the 1968 presidential election in Richard Nixon's favor.[19] Some critics argued that Shor's tweet, which was posted during the height of the George Floyd protests, could be interpreted as criticism of the Black Lives Matter movement.[20] Jonathan Chait wrote in New York Magazine that "At least some employees and clients on Civis Analytics complained that Shor’s tweet threatened their safety."[21] Shor apologized for the tweet on May 29, and he was fired from Civis Analytics a few days later.

Here's the Tweet that Shor's colleagues felt threatened their safety:

> Post-MLK-assasination [sic] race riots reduced Democratic vote share in surrounding counties by 2%, which was enough to tip the 1968 election to Nixon. Non-violent protests increase Dem vote, mainly by encouraging warm elite discourse and media coverage. http://omarwasow.com/Protests_on_Voting.pdf

These events were also covered by other media outlets. Like the Wikipedia links above, you can easily find these via Google or any other popular Internet search engine.

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

#674

Post by belorn » Sat, Feb 04, 2023, 3:53 PM UTC

notahacker wrote: post #476

riatin wrote: post #399

Earlier quoted context omitted.

How would you test for bias more effectively?

That's a good question But an even better one would be "where would you set your parameters for absence of bias" with this test I mean, take 6,774 sentences expressing negative sentiments about "gay people". I'm guessing that you're familiar with the fact that a lot of people do write these sentences, and many of them are utterly dead serious about it and genuinely do hate or at least feel a certain amount of contemp…

Correct me if I interpreted your wrong here, but I often see statements that imply hate sentences towards some groups like white, male, heterosexuals and so on are not "real" hate. The implication is that those are just ironical comparisons, jokes, or tropes.

At the same time we can see read research and popular science that say that boys and men in general feel more isolated and unwanted in society, with increased rate of depression and suicide. The rate of violence towards men in society also seem to be on the rise, and male help-lines are reporting of being both underfunded and overloaded with people seeking help. It very fair from being a joke and the consequences are very much real.

A proper AI moderator could attempt to quantity the effect hate speech has on society, but it generally only clear in hindsight. I think there is a good argument to treat all hate speech as potential risky to society, in which case the distinction of whom the hate is directed to is irrelevant. Hate is hate. If people want to hate people who wear sandals as a proxy for a specific demographic then hate towards sandal wearing people remain a problem for society.

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

#675

Post by SpeedilyDamage » Sat, Feb 04, 2023, 4:53 PM UTC

throwaway894345 wrote: post #672

SpeedilyDamage wrote: post #664

Earlier quoted context omitted.

…what do you think the title of this submission is?

Oh dear, you think every comment on HN is a direct reply to the title? No, comments are arranged in a tree structure. For example, the comment you're reading right now is a reply to your previous comment, not to the title. It's not just HN that structures its comments this way--it's a very popular schema, so this bit of knowledge should be really helpful for you on your Internet adventures. Godspeed!

Hm, so you don’t think comments should be related to one another?

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

#676

Post by SpeedilyDamage » Sat, Feb 04, 2023, 4:55 PM UTC

throwaway894345 wrote: post #673

SpeedilyDamage wrote: post #665

Earlier quoted context omitted.

I’ve said multiple times I cannot find the controversy you keep claiming happened to those two people, which to me speaks more loudly than the actual controversy they supposedly had.

I pointed you multiple times to their Wikipedia pages. Here are some direct links to the specific relevant sentences (I didn't post them before because I was busy and on mobile and I didn't realize how burdensome you would find it to find and skim their wikipedia pages). I'll quote the relevant sections as well in case you have a hard time working those links. https://en.wikipedia.org/wiki/Lee_Fang#:~:text=In%20June%…

So nothing at all to do with conservative voices being silenced, then. Why are we talking about them? You were asked for conservatives who were being silenced for saying conservative things, not generic times people got in trouble for saying things.

Kind of odd you equate racist statements with conservative statements. Why would that be?

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

#677

Post by xwolfi » Sat, Feb 04, 2023, 6:52 PM UTC

greybox wrote: post #128

prometheus76 wrote: post #97

Earlier quoted context omitted.

I think the whole point of this is that the first part of your "vs" statement and your second part can have huge overlap if you happen to agree with the worldview that is informing the responses of the AI. It looks fair to you if it spits out the response that fits your view of the world. But the whole point is, there's no such thing as "objective", especially when you start talking about people. There's no place to…

> there's no such thing as "objective", especially when you start talking about people I agree with you. However I would add to your statement about AI being revealing. One thing that AI I feel is revealing to me at least, is that it's now making me think much more about the distinction "a human search for truth" however flawed that may be, and things that are made simply to appeal to my biases. Linking this back to…

You dont have to go that far, just start from two words: reality and truth. Reality is what is, and truth is how we describe it.

You can see how truth starts to shake right ? It is an intermediary we use to interface with reality but it is imperfect: it uses eyes and ears, languages and words, expression and cultural habits, ink and paper.

For instance, I think it is true that women cannot do all than men can and men cannot do all that women can. It sounds terrifyingly horrible but because it's too hard for me to express reality. I could continue and say the differences are irrelevant to the weight of their voices that should be equal, but then... why ? I used should here, and leave reality to start ideology. I could say that opportunities should be offered in equal proportion, but... how ? Are women/men in that context two discrete categories, or a gradient where people can choose what they want to be? And do they choose wisely at all time ? What is wise ? Is it constant in all time and place ? Should we distinguish men and women ? Should we ignore what makes them different ?

And now, I trapped the AI: it can't care, it doesn't have a daughter, a mom, a wife and a sister like I do so it will not overcompensate patriarchy like I do, maybe, while I really want my daughter to prevail maybe at the cost of someone else's son, so my fair is not their fair...

Searching for truth is important, but once you've reached it, all you have is ink on a paper transcribing in a grammar and syntax what your poor eyes and weak ears perceived of reality, or at least what your ever aging brain remembers of it. It is no different for the AI, limited just as us, but not exactly overlapping the same flaws.

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

#678

Post by ackfoobar » Sat, Feb 04, 2023, 7:26 PM UTC

minsc_and_boo wrote: post #648

throwaway894345 wrote: post #644

Earlier quoted context omitted.

Of course, those are two completely different statements, as I'm sure you know: > you don't see women in engineering and leadership positions is due to them being biologically incapable. > I’m simply stating that the distribution of preferences and abilities of men and women differ in part due to biological causes and that these differences may explain why we don’t see equal representation of women in tech and leader…

Hiding behind an appeal to averages does not change the fact that Damore said women [on average ] are biologically incapable of being in tech and leadership positions. Again, Damore reaped the consequences of his own actions by violating a contract he agreed to. This is intentionally misleading to represent him as a martyr for right-wingers. Unless, of course, you think the conservative agenda includes promoting that…

Damore's argument looks more like this:

An average man is incapable of being in tech. An average woman is incapable of being in tech.

There are men and women at the tail who are capable of being in tech.

Because of the difference in averages, there are fewer women than men who are capable and want to be in tech.

---

Collapsing all that to "women [on average] are biologically incapable of being in tech" is disingenuous. Using words like "incapable" implies binariness, ignoring the continuous nature of distributions.

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

#679

Post by throwaway894345 » Sat, Feb 04, 2023, 9:24 PM UTC

SpeedilyDamage wrote: post #675

throwaway894345 wrote: post #672

Earlier quoted context omitted.

Oh dear, you think every comment on HN is a direct reply to the title? No, comments are arranged in a tree structure. For example, the comment you're reading right now is a reply to your previous comment, not to the title. It's not just HN that structures its comments this way--it's a very popular schema, so this bit of knowledge should be really helpful for you on your Internet adventures. Godspeed!

Hm, so you don’t think comments should be related to one another?

The comments were related to each other, and indirectly to the title. This discussion has descended into obtuseness, and I find obtuseness to be boring, so I’ll dip out.

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

#680

Post by throwaway894345 » Sat, Feb 04, 2023, 9:31 PM UTC

SpeedilyDamage wrote: post #676

throwaway894345 wrote: post #673

Earlier quoted context omitted.

I pointed you multiple times to their Wikipedia pages. Here are some direct links to the specific relevant sentences (I didn't post them before because I was busy and on mobile and I didn't realize how burdensome you would find it to find and skim their wikipedia pages). I'll quote the relevant sections as well in case you have a hard time working those links. https://en.wikipedia.org/wiki/Lee_Fang#:~:text=In%20June%…

So nothing at all to do with conservative voices being silenced, then. Why are we talking about them? You were asked for conservatives who were being silenced for saying conservative things, not generic times people got in trouble for saying things. Kind of odd you equate racist statements with conservative statements. Why would that be?

We’ve already been over this. I wasn’t asked for conservatives who were being silenced—I was refuting the argument that the people being canceled or “called out” were saying reprehensible stuff. I certainly never equated racist statements with conservative statements. It seems like you’re not following the context of the thread and that’s adding a lot of confusion—clarifying basic thread context is not my idea of an interesting debate (not to mention repeating the same conversation about thread context) so I’m probably going to dip out.