Hilariously enough, 4chan, if it is still functions how I remember it, sounds like it would be the most protected given the lack of algorithmicly promoted content.
Still, trying to prove a negative though, isn't it?
After reading on /pol/ horrible takes on Jewish slurs, and horrible takes on BiPOC folk, and horrible takes on homosexuals.... And it's the 1st post...
The OP's point is that since the case hinges on the question of whether social media platforms are "mere messageboards," or if they are, "products," the presence of hate content (like you'd find inside the telephone switching network) and the absence of any kind of algorithmic management is helpful to Nishimura, while the complex recommendation systems on YouTube will serve as an obstacle for that defendant.
Moderation as censorship typically does not cater to, or curate content for, distinct individual users, let alone audience groups. There are of course exceptions but this is about a different extreme.
This all started with ads. Now all content is an ad, so that you can be served more ads. It's ingenious but also evil.
Threatening to "go kill ni**ers for a high score" is likely illegal.
That shits on /pol/ right now, and is the root of supporting extremism to the point of violence and murder. And last I checked, making death threats is explicitly not protected speech.
Your test sentence fails the Brandenburg test: it is not "directed to inciting or producing imminent lawless action", nor "likely to incite or produce such action."
However, "go kill ni*ers for a high score" is not only not a true threat, it's the kind of speech that Brandenburg v Ohio, which established the current incitement-to-imminent-lawless-action test, explicitly said was 100% protected free speech.
Pretty much anything less than leading a mob and telling them who is their next lynching target will be ruled protected speech.
From reading the article, it sounds like what the judge actually ruled is not what the headline seems to claim that the judge ruled. It seems that the judge is simply saying the case can proceed to the next phase. The plaintiffs, it appears, still need to argue their claim of product liability. So they still need to present a case to the court that each of the defendants products are in fact defective.
I wonder if this is "the disconnection of article and headline" sort of thing. I've read some article authors have no control over the headlines associated with their writings, either added by someone else, or generated or a/b tested or whatever.
I think that's more because software is in everything and the product that software is in is what gets the liability as a whole.
iirc, the Toyota breaking issue & lawsuit was ultimately a software issue. The code was such a mess that I decided I wouldn't buy Toyota again. This was maybe 8-10 years ago
And so begins another round of tech company funded shill outlets (hi Mike) and lobbyists convincing everyone that the most valuable companies on the planet must be given absolute immunity for their behavior or else the Internet will collapse... even though that immunity law is only in the US anyways.
Would be interesting to see a similar headline like: "Judge demands FBI prove they didn't help radicalize mass shooter", which is definitely fairly common if measured as a % of terrorist convictions.
More to the point of the article: I don't understand all these attacks on Section 230. Is there a single reasonable proposal for what kind of framework should replace Section 230?
The headline is simply a clickbait lie (fairly typical of the source); what the the judge actually ruled is not that liability is assumed and platforms must disprove it, but that Section 230 is not a bar to a defective product liability claim.
The problem with that is that the only "product" here is their website.
Regulating harmful speech as a "product" seems a bit odd because the product exists only to convey certain speech. Any device used to convey harmful speech would seem to become subject to liability under this theory, as all claims of how this caused harm seem to be related to conveying speech that lead to harmful beliefs, even if that is dressed up as some product that is a "sophisticated system" for recommending or promoting bad ideas.
This becomes an especially thin fig leaf when applied to, e.g., 4chan which doesn't seem to do anything like recommending or promoting certain speakers as far as I can tell.
Calling something a website isn't descriptive enough, plenty of websites get shutdown so being a website alone isn't enough to qualify as protected speech.
4chan and Facebook both are already filtering out Child Porn, they may need to add other categories... or not it's an ongoing question.
More abstractly I think we can agree a phone network or ISP is a less active participant in what specifically gets said than a recommendation engine. So an websites internal recommendation engine will ultimately get its own kinds of regulations eventually just like search engines etc.
True threats are a distinct category that may sometimes overlap with hate speech. But those statements are illegal because they threaten immanent lawless action, not because they're hateful.
True threats are someone saying the speaker will do something that’s believable. Ie, Someone claiming they are going to Nuke DC without a nuke fails the standard. Someone claiming they’re going to shoot you while holding a gun is.
So it can be illegal hate speech by inciting people to act without being classified as a true threat. The triggers being inciting, specific, and immediate aren’t on their own illegal. I can suggest people go to Starbucks and get a coffee without issue.
The part I was trying to get at is that under current law, true threats are their own category. Hate speech that is a true threat is illegal, due to being a true threat. Hate speech that is not a true threat is awful, but legal.
So hate speech is not illegal for being hate speech, it's only illegal when it's also a true threat. And in that case it's illegal because it's a true threat.
There's certainly plenty of overlap, but it doesn't help anyone's understanding of why it's illegal to lump them together or point out that speech can fit in more than one category when only one of the categories is illegal under current US jurisprudence.
I'm fairly certain that any two people going to the same page at the same time on HN will see the same thing. What they see will mostly be determined by the age of a post/article along with up/down votes by end users. It would be difficult to argue that much is going on here beyond speech.
Yet a site like YouTube is different. It will display different content to different people based upon their history (and, likely, other factors). You don't even have to be logged into YouTube for this to happen. It is a lot more difficult to argue that this is a form of speech since the recommendations are most likely the product of some optimization algorithm. It certainly isn't the product of the direct input of other people.
> Would that not depend upon how the website works?
They're suing, among others, 4chan which is the same as HN in that regard and the judge let the lawsuit go forward.
So whatever might later be litigated, we can see from this that the plaintiffs currently have a "defective product" theory that HN would also fit if only there were some murderer willing to claim that they were somehow radicalized by HN.
The defendants (4chan and the others) are independent. It's possible 4chan will be found not to be liable (no editorial control exercised in such a way as to create new speech) and the others liable.
Note that the judge did not find that any of the websites are products or subject to product liability, only that there is a justiciable dispute of fact on that point. Your argument hinges on facts that are on dispute and on which evidence will be presented later by the parties, at this stage the standard is does the complaint state a cause of action when viewing every disputed fact in the light most favorable to the plaintiffs. This is not a ruling on the facts, it is a ruling on whether the filing raises issues for which it is necessary to examine the facts to reach a resolution.
No, Section 230 does not require providers to not exercise editorial control. This is a commonly repeated misunderstanding of the law. There is no room for a court to find GP's misstatement in the existing law.
§230 is one of the most misunderstood laws in the US, in large part because of intentional efforts to misconstrue it. What it actually says is that websites are not to be considered the speaker of any user-generated content, and there is explicitly and very intentionally no precondition along the lines of "publisher-platform" or "algorithmic content" or "moderation" or the like.
Now, it's also the case that moderation and content recommendations themselves are speech on the part of the website--crucially so, for that means they also have First Amendment rights to choose those policies without government coercion. There is an interesting question as to whether it is possible for such speech to create liability independent of the underlying user content (for which §230 absolutely immunizes them); indeed SCOTUS heard a case not long ago to answer this question, only to toss it out because the plaintiffs were uninterested in discussing that question.
The interesting question is likely (IMO) to be that it can create liability.
Selectively picking & ordering the content of others can trivially create new meaning, for example the classic trope of a ransom note formed by clipping words & phrases out of a newspaper & gluing them onto a page. Whether a particular site's editorial actions on user content create liability is likely to be determined on a case-by-case basis. Real site's algorithmic recommendation systems aren't nearly as transformative as newspaper clipping ransom notes, but the sequence and organization of recommended content may well count as a new composite work (much like a music album is a work composed of a sequence of individual tracks) and create liability. I wonder if it would also qualify for individual copyright.
Section 230 is being abused by criminal companies who want to have their cake and eat it too.
They act as a publisher and algorithmically tailor content to each individual user to increase engagement and still enjoy the protection of S230. Generally this results in noxious, harmful and/or illegal content being pushed to the top (as outrage or indignation yields way more engagement).
This is far from what S230 was originally meant to protect - content-agnostic communication mediums where the provider is simply facilitating the exchange of user-generated-content without exerting editorial control outside of edge-cases such as spam filtering/moderation.
I think there's an interesting line to consider around 'engagement algorithms' being responsible for accelerating radicalization (i.e. seeing certain content and then push notifications of same/similar content get delivered.)
The 3rd rail of course, is that to fix that you would likely either enter a 'culture war' with each side challenging the others, or you just have to go back to a non-engagement tuned way to deliver updates [0].
And, you know, it's worth some sort of discussion, because I've seen that one mis-click of a youtube video on my phone (trash screen protector) or inebriated TikTok scroll session where it takes me too long to realize it was one form or another of content I didn't really care for before switching on, and then it can take up to a month before the pop ups suggesting similar content stop.
But, if the algorithm is 'fair' internally, I would guess the content provider will argue that it is such and, well, sucks this person didn't click a 'save the whales' link instead?
To be clear, this is why it is worth a conversation, but it is a difficult one for society for a number of reasons. However I do miss the charm of the 'old web' where we relied on word of mouth. I think MySpace really was a certain peak of Minimalistic non algo Social media; you picked your sort for who to pay attention to, you could put your links or whatever, toss a song for a mood, and it was simple.
On the flip side, something like 4chan is not something, last I checked, that did engagement or anything past 'comment = bump' (-does- sage even work on it? did it ever?). AFAIK the argument would have to rest to some extent on proving that the mods failed to 'properly' follow 230... extreme example where they could be in trouble in my mind would be 'a post was reported for a legitimate violent threat and instead of removing/locking mods pinned it'.
[0] - you know, when they were nice 'rollups' of what happened more like your mail vs the modern 'spam' of each thing, to get you back in the app and ideally tap each one to view so somebody meets an OKR.
>However I do miss the charm of the 'old web' where we relied on word of mouth. I think MySpace really was a certain peak of Minimalistic non algo Social media; you picked your sort for who to pay attention to, you could put your links or whatever, toss a song for a mood, and it was simple.
This is because back then you only had yourself to rely on to find the content you wanted.
Now everybody has to use their algorithms to predict what you want to see, which is terrible.
What I've noticed in the past couple years is that on any social platform that I admittedly choose to use, all it really takes is clicking something for one topic maybe once or twice and then half my feed is constantly full of it with no way to say "no that was a one-time thing".
Even YouTube is bad at this. Think you clicked a game or programming tutorial article and it is actually a link to YouTube? Better not watch if you don't like having your recommendations cluttered with videos you'll never be interested in.
At least with YouTube you can delete items from your watch history, but it's a few clicks to get to, so especially annoying for those one-off videos.
I really, really wish there was a slightly higher threshold stopping those one-time mistakes from becoming a time sink.
> The 3rd rail of course, is that to fix that you would likely either enter a 'culture war' with each side challenging the others, or you just have to go back to a non-engagement tuned way to deliver updates
There is another option, to give users choice and control over the algorithms and content moderation. Bluesky had a recent blog post outlining their Stacked Moderation scheme and open sources project that allows anyone to create a moderation server, which you can use in Bluesky. It would seem this ought to apply to the feeds as well, which they seem to offer a better paradigm than Twitter etc Al already, but not yet what I'm talking about
I think there is a difference between a message board where everyone sees the same messages and surveillance software which spy on users to build behavioral profiles, which are then used to tailor content for each user. The surveillance software is an active participant in the conversation. It is no longer merely the medium.
I think it’s helpful to not say things like “surveillance software”, it’s too easy to have those words become a distraction.
When you sort by date you should have quite a lot more immunity than when you pick the content a person sees. Choosing content can manipulate a person and as a result start sharing the responsibility for their actions.
Where on the spectrum between protected first amendment expression and conspiracy does a newsfeed algorithm lie? This is a new question.
But it's so funny that the CIA-funded USG spyware doesn't want to let anyone influence what they show their users! All the more so for the TikTok furor.
"How dare you propagandize to our citizens! That's our job and turf!"
Won’t somebody think of the children. All communication must be filtered through benevolent feed algorithms! Oh no the benevolent feed algorithms are creating pedophiles and terrorists! I yearn for the days of unfiltered chronological communications. Won’t someone think of the children! All communication must be curated by machine learning algorithm. Oh no the machine has created monsters! Wasn’t the web better without the algorithms? Think of the children! We must censor the web! Oh no the web sucks and we still have pedophiles and terrorists. Wouldn’t it be better to go back to 1999? But think of the children.
There has always been a large spectrum from overly censorious forums to 4chan. There’s a lot of room in between those. I don’t want nor do I visit 4chan but I do want something less censored than Reddit.
It's more important than you're implying, because it subjects the sites to discovery, which is quite burdensome. Avoiding that sort of trap is why we have SLAPP laws.
It's definitely true that the defendants would have preferred to have the case dismissed, even if they will win eventually (the process is painful and costly). But that isn't really what the clickbait headline says, is it?
Eh, it confuses what's actually stated in the ruling with the direct results of the ruling, which would be a trial on the merits of this, in which they'll have to establish by the preponderance of the evidence that they're not a link in the chain of whatever radicalized this guy.
So I can see that it's misleading about what the ruling says, but not about the net effect of the ruling. But it actually links to the actual ruling here, so it still comes out as a C+ article over all, merely because of the terrible state of legal reporting.
> Eh, it confuses what's actually stated in the ruling with the direct results of the ruling, which would be a trial on the merits of this, in which they'll have to establish by the preponderance of the evidence that they're not a link in the chain of whatever radicalized this guy.
No, they won't necessarily. As the ruling makes clear, their Section 230 immunity claim is still live, but there are disputed questions of fact nexessary to resolve it. They only need to refute whatever proof is presented for causation if the plaintiff shows sufficient eveidence that they are a product in a manner not protected by Section 230 and they fail to refute that.
I wasn't claiming that SLAPP applied, I was claiming that it's used to avoid even getting to discovery when it does apply, because even letting cases get that far is expensive.
"It's not clear when - or even if - the case will go to trial. But it's headed that way unless the platforms successfully appeal or settle the matter. The result could set a precedent that makes it harder to use Section 230 as a get-out-of-jail-free card for social media platforms - something that US officials have long tried to do."
They always settle.
Could also set a precedent that makes it _easier_ to get cases dismissed on Section 230, but "Big Tech" will settle for a larger amount than it costs to go to trial and get that result. Year after year they do this.
72 comments
[ 2.6 ms ] story [ 143 ms ] threadStill, trying to prove a negative though, isn't it?
Yeah that's going to end badly.
The difference is that posts don't even make it past the gate if they get blocked.
And certain content is also ban-worthy.
Again, they very much do censor stuff. You just don't see the censoring.
This all started with ads. Now all content is an ad, so that you can be served more ads. It's ingenious but also evil.
That shits on /pol/ right now, and is the root of supporting extremism to the point of violence and murder. And last I checked, making death threats is explicitly not protected speech.
Your test sentence fails the Brandenburg test: it is not "directed to inciting or producing imminent lawless action", nor "likely to incite or produce such action."
https://www.law.cornell.edu/wex/brandenburg_test
However, "go kill ni*ers for a high score" is not only not a true threat, it's the kind of speech that Brandenburg v Ohio, which established the current incitement-to-imminent-lawless-action test, explicitly said was 100% protected free speech.
Pretty much anything less than leading a mob and telling them who is their next lynching target will be ruled protected speech.
I think it can be beneficial for social media sites and other ventures to collaborate with society more
iirc, the Toyota breaking issue & lawsuit was ultimately a software issue. The code was such a mess that I decided I wouldn't buy Toyota again. This was maybe 8-10 years ago
More to the point of the article: I don't understand all these attacks on Section 230. Is there a single reasonable proposal for what kind of framework should replace Section 230?
Regulating harmful speech as a "product" seems a bit odd because the product exists only to convey certain speech. Any device used to convey harmful speech would seem to become subject to liability under this theory, as all claims of how this caused harm seem to be related to conveying speech that lead to harmful beliefs, even if that is dressed up as some product that is a "sophisticated system" for recommending or promoting bad ideas.
This becomes an especially thin fig leaf when applied to, e.g., 4chan which doesn't seem to do anything like recommending or promoting certain speakers as far as I can tell.
4chan and Facebook both are already filtering out Child Porn, they may need to add other categories... or not it's an ongoing question.
More abstractly I think we can agree a phone network or ISP is a less active participant in what specifically gets said than a recommendation engine. So an websites internal recommendation engine will ultimately get its own kinds of regulations eventually just like search engines etc.
So it can be illegal hate speech by inciting people to act without being classified as a true threat. The triggers being inciting, specific, and immediate aren’t on their own illegal. I can suggest people go to Starbucks and get a coffee without issue.
So hate speech is not illegal for being hate speech, it's only illegal when it's also a true threat. And in that case it's illegal because it's a true threat.
There's certainly plenty of overlap, but it doesn't help anyone's understanding of why it's illegal to lump them together or point out that speech can fit in more than one category when only one of the categories is illegal under current US jurisprudence.
I'm fairly certain that any two people going to the same page at the same time on HN will see the same thing. What they see will mostly be determined by the age of a post/article along with up/down votes by end users. It would be difficult to argue that much is going on here beyond speech.
Yet a site like YouTube is different. It will display different content to different people based upon their history (and, likely, other factors). You don't even have to be logged into YouTube for this to happen. It is a lot more difficult to argue that this is a form of speech since the recommendations are most likely the product of some optimization algorithm. It certainly isn't the product of the direct input of other people.
They're suing, among others, 4chan which is the same as HN in that regard and the judge let the lawsuit go forward.
So whatever might later be litigated, we can see from this that the plaintiffs currently have a "defective product" theory that HN would also fit if only there were some murderer willing to claim that they were somehow radicalized by HN.
[edit] I stand corrected.
Now, it's also the case that moderation and content recommendations themselves are speech on the part of the website--crucially so, for that means they also have First Amendment rights to choose those policies without government coercion. There is an interesting question as to whether it is possible for such speech to create liability independent of the underlying user content (for which §230 absolutely immunizes them); indeed SCOTUS heard a case not long ago to answer this question, only to toss it out because the plaintiffs were uninterested in discussing that question.
Selectively picking & ordering the content of others can trivially create new meaning, for example the classic trope of a ransom note formed by clipping words & phrases out of a newspaper & gluing them onto a page. Whether a particular site's editorial actions on user content create liability is likely to be determined on a case-by-case basis. Real site's algorithmic recommendation systems aren't nearly as transformative as newspaper clipping ransom notes, but the sequence and organization of recommended content may well count as a new composite work (much like a music album is a work composed of a sequence of individual tracks) and create liability. I wonder if it would also qualify for individual copyright.
Is there any hard, legal record of this? That the FBI radicalized someone?
2020: (Whitmer kidnapping plot) https://archive.is/Met7X
2023: (Humzah Mashkoor) https://thehill.com/video/fbi-agents-groomed-autistic-16-yr-...
2023: (Mateo Ventura) https://archive.is/hb9Im#selection-627.47-627.60
They act as a publisher and algorithmically tailor content to each individual user to increase engagement and still enjoy the protection of S230. Generally this results in noxious, harmful and/or illegal content being pushed to the top (as outrage or indignation yields way more engagement).
This is far from what S230 was originally meant to protect - content-agnostic communication mediums where the provider is simply facilitating the exchange of user-generated-content without exerting editorial control outside of edge-cases such as spam filtering/moderation.
The 3rd rail of course, is that to fix that you would likely either enter a 'culture war' with each side challenging the others, or you just have to go back to a non-engagement tuned way to deliver updates [0].
And, you know, it's worth some sort of discussion, because I've seen that one mis-click of a youtube video on my phone (trash screen protector) or inebriated TikTok scroll session where it takes me too long to realize it was one form or another of content I didn't really care for before switching on, and then it can take up to a month before the pop ups suggesting similar content stop.
But, if the algorithm is 'fair' internally, I would guess the content provider will argue that it is such and, well, sucks this person didn't click a 'save the whales' link instead?
To be clear, this is why it is worth a conversation, but it is a difficult one for society for a number of reasons. However I do miss the charm of the 'old web' where we relied on word of mouth. I think MySpace really was a certain peak of Minimalistic non algo Social media; you picked your sort for who to pay attention to, you could put your links or whatever, toss a song for a mood, and it was simple.
On the flip side, something like 4chan is not something, last I checked, that did engagement or anything past 'comment = bump' (-does- sage even work on it? did it ever?). AFAIK the argument would have to rest to some extent on proving that the mods failed to 'properly' follow 230... extreme example where they could be in trouble in my mind would be 'a post was reported for a legitimate violent threat and instead of removing/locking mods pinned it'.
[0] - you know, when they were nice 'rollups' of what happened more like your mail vs the modern 'spam' of each thing, to get you back in the app and ideally tap each one to view so somebody meets an OKR.
This is because back then you only had yourself to rely on to find the content you wanted.
Now everybody has to use their algorithms to predict what you want to see, which is terrible.
What I've noticed in the past couple years is that on any social platform that I admittedly choose to use, all it really takes is clicking something for one topic maybe once or twice and then half my feed is constantly full of it with no way to say "no that was a one-time thing".
At least with YouTube you can delete items from your watch history, but it's a few clicks to get to, so especially annoying for those one-off videos.
I really, really wish there was a slightly higher threshold stopping those one-time mistakes from becoming a time sink.
There is another option, to give users choice and control over the algorithms and content moderation. Bluesky had a recent blog post outlining their Stacked Moderation scheme and open sources project that allows anyone to create a moderation server, which you can use in Bluesky. It would seem this ought to apply to the feeds as well, which they seem to offer a better paradigm than Twitter etc Al already, but not yet what I'm talking about
When you sort by date you should have quite a lot more immunity than when you pick the content a person sees. Choosing content can manipulate a person and as a result start sharing the responsibility for their actions.
Where on the spectrum between protected first amendment expression and conspiracy does a newsfeed algorithm lie? This is a new question.
"How dare you propagandize to our citizens! That's our job and turf!"
Maximum stack size exceeded exception.
So I can see that it's misleading about what the ruling says, but not about the net effect of the ruling. But it actually links to the actual ruling here, so it still comes out as a C+ article over all, merely because of the terrible state of legal reporting.
No, they won't necessarily. As the ruling makes clear, their Section 230 immunity claim is still live, but there are disputed questions of fact nexessary to resolve it. They only need to refute whatever proof is presented for causation if the plaintiff shows sufficient eveidence that they are a product in a manner not protected by Section 230 and they fail to refute that.
SLAPP is irrelevant here, since this was not an anti-SLAPP motion to dismiss, it was a motion to dismiss for failure to state a claim.
They always settle.
Could also set a precedent that makes it _easier_ to get cases dismissed on Section 230, but "Big Tech" will settle for a larger amount than it costs to go to trial and get that result. Year after year they do this.