Supreme Court Doesn't Want to Create ‘World of Lawsuits’ for Tech Firms
bloomberg.com
bloomberg.com
I've read a couple of articles covering this court case, and the algorithms are talked about, justices have wondered if they are neutral or not. But nobody has asked "What is the algorithm?"
We don't know. It could be code that makes recommendations based on similar users, or it could be someone in a basement picking what to push on people based on some agenda, or the algorithm can use any of those hundreds of advertising categories they sort people into - show white people this video, married people this video, people in this country something else, these voters something over here, make exceptions for famous people or the government or the establishment, what some rich person paid us to push, etc. These algorithms affect everything from pop culture to elections to genocides to school bullies and teenage suicides.
We have no idea what these big tech companies are doing, except that they are the new mass media, they don't want to accept any responsibility for their actions, and profit is more important to them than social externalities.
What I don’t understand is: what does this have to do with Section 230 and with the Supreme Court. Section 230 exempts firms (correctly) from liability for user content even if they moderate content in various ways. It doesn’t exclude complex algorithms because those algorithms weren’t invented at the time the law was written. If people want to write new laws, they should contact their politicians and get them to do that, stop trying to add new requirements to a law that didn’t consider them.
I don't know exactly how to write such a law, but I like the concept of treating an algorithm that provides personalized, maximally addictive content as the publisher of that content. Don't want to get sued? Curate your content (Netflix) or don't use that kind of algorithm (reddit).
I’m not saying you’re wrong on the law. I’m saying that it feels obvious that the goal here is coercion. While I suppose some people might legitimately want to ban all use of algorithms (including moderation?) it seems more that the idea is to “work around” the spirit (and perhaps literal text) of the First Amendment by using the courts to make non-compliant publishers unprofitable. At very least it feels like “Congress passing laws abridging” if not literally banning.
I also don’t understand the limits of this approach. Are there any principles here? If some near-future Congress passed a law saying “any algorithms you use must reduce the reach of posts that criticize the Biden administration, else you major platforms lose liability protection,” would that be fine? In my eyes that seems very abusive and problematic.
It seems to me the guiding principle should be that a sufficiently smart, personalized algorithm gets treated like a human editor.
Of course, precisely defining "sufficiently smart" in a manner that's both reasonable and enforceable might be hard.
As far as the actual text of the First Amendment goes, I’m not aware of any distinction between platforms and publishers: there’s simply a broad prohibition on Congress passing laws that abridge freedom of expression.
And let’s be clear: abridging the speech rights of these private parties is precisely the intention of many lawmakers who have proposed Section 230/liability updates, both at the state and Federal level. They’re mad because they think private companies are suppressing content that they want published, and they want to use the power of the state to punish such infractions. Note that they don’t want to punish the companies for the specific moderation decisions they object to, since that would obviously be protected speech. No: the political goal here is to punish private firms who violate government speech codes by causing them to suffer for unrelated posts that appear on the platform, as a kind of protection racket. Nobody is even being coy about this stuff.
A publisher selects or whitelists content. A platform does not, but may moderate or blacklist content. Of course, mathematically those can achieve the same result, but laws are not mathematical formulas. The intent and effects matter.
I'm aware some of the proposals to update platform immunity are intended as political leverage. I oppose those.
Not to be the "citation needed" person, but can you point me to the specific legal statute that makes this distinction -- i.e., says that platforms can only blacklist content? Or are you saying that you think this should be the definition and we should pass new laws that enforce limits on what platforms can do?
These companies are selecting and amplifying certain speech. The fact that the process is somewhat automated doesn't change this. If any legal argument relies on "the algorithm" being likened to simple searching and sorting, maybe we should require some transparency there so anyone can know if that's actually the case.
On another note, Section 230 isn't a free pass to host illegal content. It doesn't make a website immune to DMCA takedown requests, for example. We could have terrorism takedown requests, and the burden wouldn't harm large, irresponsible companies or small, moderated forums.
There is no algorithm.
Rather, there are tons of algorithms. Hear me out...
If you really want to what a company is doing, then you have to go back to the data ingesting process:
- How do they scrape?
- How to they choose what to ignore?
- How often is something scraped?
- How is the scrape timing determined (e.g. is that yet another algorithm?)
- What technology do you use to verify that the scraped data is the real site (e.g. MITM, hacked sites, etc)
- How do you determine if data should be removed?
- Is data removed or just hidden?
- Do you have any sites that you blacklist (i.e. don't scrape)?
- Do any governments require you to either scrape or not scrape certain sites/pages/etc.
These are just some of the questions that I would want to understand. For instance, if I'm an evil guy, I'm only going to scrape evil sites, and then I can use the same "algorithm" everyone else uses, but with vastly different results because my ingested corpus is so different.
If we are only focusing on "the algorithm", then we've already lost, because we don't truly understand the scale of the problem, - and most importantly - we definitely don't understand how to regulate it.
I think that the user content recommendation algorithms might be (or could be!) easier to describe than search page-rank algorithms, not needing to worry about some of the things you mentioned like scraping (because content is posted by users). Other things you mentioned are still relevant, like blacklists and ignoring content, now that these are platforms are so influential, often being referred to as the new public square.
Why can't facebook let you see posts from your friends in chronological order, or geographically close, or some other set of understandable options? Why is facebook protected when they inject unsolicited posts about something outrageous happening, just to keep you online, even if it's something harmful? How does facebook choose which posts to put in front of you, contnent that you did not ask for? Why does youtube recommend flat earth videos so often? How many big override knobs does twitter have to decide what goes viral, and what does not? The same for tiktok. Who at facebook and youtube and twitter and tiktok gets to decide this stuff for the billions of us using their platforms?
I haven't thought as much about Section 230 as it applies more specifically to something like google search and page-rank. That is a related topic. But at least you have to ask for something on a search engine, it doesn't bait you with unrelated content like the recommendation engines on the platforms I mentioned. I do agree with you that this can get complicated.
These recommendation algorithms are very influential - this is the new media, much bigger than anything before. If they can't describe how or why they promoted a harmful piece of content in a trial, then perhaps they should not be promoting content? Otherwise we are being manipulated on an unprecedented scale by a few insiders with very little oversight.
Unfortunately, flicking through all the briefs, it seems that no one actually wants to answer that question. Instead, everybody is trying to argue for or against §230 more or less in its entirety. I guess people took Thomas's complaint about §230 in a prior dissent and everyone assumed they could find enough votes on SCOTUS to gut §230 entirely? (Ironically, Thomas seems to be firmly pro-§230 in this specific case).
For example, NBC was sued by GM over a rigged SUV crash video. Fox News is being sued by Dominion for defamation. There are other examples. TV networs don't have legal immunity.
Instead of aggressively monetizing "engagement" with advertising via physiological engineering and manipulation. Matching each person to the most addictive content, customized to that person's mental weaknesses.
I.e. creating and leveraging mental health issues is their goal.
1. The value of a social media site is directly proportional to the number of people you want to engage with who are on the same site, and hence directly proportional (probabilistically) to the total number of users. A free site is always going to attract more users than a paid one, so a paid site will always be at a competitive disadvantage. Because social media tends to be a winner-take-all economy, no paid site will ever survive.
2. Different people want and need different things from a social media site, and so no one policy is ever going to make everybody happy. For example, the needs of people who want free speech will always be in conflict from the needs of people who want to filter out falsehoods.
Other than that, it's a cake walk.
> charged a small monthly fee
These two goals are diametrically opposed. Social media sites only hit critical mass when everyone has free access to them.
But I don't see how you can argue the algorithm is entirely neutral, if the algorithm is to drive engagement. The algorithm is not recommending a random video with 'cooking' in its description. It's going from 'cooking', to 'eat more vegetables', to 'have you considered vegetarianism,' eventually to 'vegan purity must be maintained - meat eaters must die' and 'we gloriously behead a poultry farmer.' At some point, when you're making decisions about what to recommend, you're no longer just passively hosting content.
Even if the algorithm wasn't intentionally written to turn people into a radical vegetablists, at some point it shouldn't matter. And (again, my recollection) Google didn't jump up and say "Gadzooks! You're right! We shall cease immediately for we are not evil (despite our handle-bar mustaches)" They, and other social content companies, fought the changes. So it's not as if they were ignorant of what they were doing.
Meh - I think some of the other posters are right - they'll kick it down the road and hope it goes away.
CompuServe didn't moderate at all, while Prodigy did moderate. The courts found Prodigy liable for content, because it chose to moderate some content. CompuServe was not liable because it left all content up. Congress wanted to encourage sites to moderate content without suddenly becoming liable for anything they failed to moderate, so 230 exists to shield sites who choose to moderate some content from becoming liable for all the other content that they fail to moderate in good faith (https://www.eff.org/issues/cda230/legislative-history).
§230 is explicitly pretty broad on immunity: it states that providers are neither liable for failing to moderate nor for the consequences of their moderation. "Good faith" is a requirement, but only for actions taken to restrict access to the material (i.e., it doesn't apply to a failure to moderate.)
Section 230 doesn't require that platforms be neutral.
> “If it’s the same algorithm to present cooking videos to people who are interested in cooking and ISIS videos to people who are interested in ISIS, racing videos to people who are interested in racing, I think you’re going to have to explain more clearly if it’s neutral in that way how your claim is set apart from that,” Thomas told Schnapper.
I don't think it's just a matter of representing racing videos to people interested in racing. For example, I start with a video on police misconduct and before I realize it, I'm seeing videos telling me as a sovereign citizen I can can't be arrested ... because magic.
In point of fact there probably is some stuff to worry about, but where we've seen it it's been almost exclusively in off label sites. Unpopular opinion: the core tech portals have been remarkably good stewards of the public good in the face of genuine extremis in the discourse the public wants to engage in.
Everyone's got their favorite Moment of Clear Censorship that Must Not Be Permitted, but name any forum and any topic, and I'll be able to show you (with a quick seearch using tools provided by these very tech giants) extensive unimpeded discussion.
To wit: SCOTUS isn't going to step in because one tweet got taken down. There are thousands just like it that weren't censored.
You see it in crowds, for example, when people go from having one energetic but good time to being put in a state where they are likely to commit violence. I don't know if Travis Scott is more legend than phenomena, or if the Woodstock '99 people realized they created a metal festival. Human beings are weird creatures and we can do horrific things, given the right context and environment. I think the actual text is rarely a cause for concern, but the manipulation around it is. You can take the same group of people and have an orderly discussion on defenestration, but work them up and you can get an actual defenestration.
So if a company is manipulating the flow of information to create an emotional state, on purpose, I'm not sure that company should be 100% in the clear on it. Especially since the law seems to be more about good faith efforts at restricting access to content.
This creates several new business models and opportunities for improvement
YouTube, et al, simply ingests, stores, and provides videos with extensive metadata profiles for query, but without any search or editorial/optimization algorithms
A separate business layer of 'channels' or 'networks' provide curated content from the host database business using the metadata and user preferences.
Imagine the 'wild animal' channel, or 'cool videos from the 70s' channel
The channel layer survives by meeting the market segment demands and received subscription fees or advertising revenue
The host also can also inject ads or charge subscription fees for access to the metadata for channel queries.
Neither layer violates section 320
Separating data hosting and curation creates an entirely new media market and opportunities for consumer choice
Like the phone and airline monopoly breakups, I suspect this change requires legislative action
I read the lawsuit (it's on the Supreme Court website). A lot of those questions relate to arguments being made by the plaintiff about how to let them win without destroying Section 230 entirely.
So, this is my prediction: they will "win" because the Supreme Court is going to "clarify" the definition of terms in Section 230. It will not have much of any impact on sites like HN or Reddit, but Google/MS will have some real issues unless... They go all-in on LLMs. But they had better get them right, because those things have zero protection. Oh, and FB/Twitter are toast.
Get ready, things will be interesting...
Not to say it couldn't happen. But this is the branch of government least fond of getting attention.
In the meantime, we have small sites making users happy in select areas.
And massive site driving the world crazy whose business model is find the most addict-able content possible for every human being.
I.e. creating a mental health issue is their goal.
From my reading of Gonzalez, it's strictly the latter kind. The good part of that is that whatever the outcome Congress could turn around tomorrow and overrule it with new legislation so it really doesn't matter.
Google in particular has done a very good job of painting things like recommendations and search result ranking as "oh it's the algorithm". But who coded the algorithm? As we saw with the Panda update [1] targeting low-quality content farms, subjectivity comes into the algortihm in deciding what is and isn't good content.
Ultimately this case is deciding if the result of a recommendation algorithm represents content in its own right. I imagine a sbuejctive test will come out of this case that will be fairly narrow and match what companies are doing anyway. The CDA requires this already. There has to be a process to report and takedown content.
When looking at how the Supreme Court rules on cases you can pretty reliably answer that from a materialistic perspective. Opening up tech companies to a conflagration of litigation is of no benefit to corporations so that probably won't happen.
In this case, the Anti-Terrorism Act refers to "international terrorism". That's an interesting carve out for domestic terrorism like, say, mass shootings, attacking infrastructure like power stations, bombing abortion clinics or, you know, insurrection.
It is worth having a discussion about the biases that creep into recommendation algorithms because they're quite real and we've created a pipeline for radicalizing people sometimes into violence (eg Alexandre Bissonnette, Scott Roeder [2], Payton Gendron [3]).
[1]: https://www.searchenginejournal.com/google-algorithm-history...
[2]: https://www.rollingstone.com/politics/politics-features/bill...
[3]: https://www.nbcnews.com/news/us-news/fox-news-tucker-carlson...
They could, but they won't. We currently have a divided Congress, and neither house will pass something that the other would be willing to tolerate. Even if both houses and the President were of the same party, it almost certainly wouldn't survive a filibuster in the Senate.
It's a political issue -- and all issues are political issues -- so Congress won't act in a way that one party could see as a loss for them. Congress hasn't passed a nontrivial law in years and they won't start with this.
Good, but let's remember that it's still their property, they just let users use it, according to their own rules. The protection from lawsuits gives the tech giants more power with less responsibility.
I believe the world would be a better place if web properties took more responsibility of what happens on their premises. The internet would skew back to smaller web properties, and a centralized media giant would not be able to close anyone down because they don't like them.
Tech savvy people would make their own web properties, others would use it. It may or may not be connected by a federated communication protocols, but the important piece is that they own the property and be responsible of what happens there. E.g. if you invited your grandma to your forum -- be responsible of what she writes there.
PS: Of course, illegal content can still be closed down (e.g. through hosting providers and domain registrars), but not by a whim of a private corporation.
It seems more likely to me that increased legal liability on content hosting platforms would have exactly the opposite effect. Smaller platforms aren't going to have the resources or desire to deal with the legal headache of endless lawsuits because someone didn't like what one of their posters had to say, so they're either not going to operate at all or ban everything even remotely offensive to avoid legal problems. And the larger platforms you're concerned about would have even more power, since only they can afford to play the legal liability game, and even more incentive to ban controversial speech to mitigate those legal concerns.
Sell the ability to host privately available content, even social media. The content is not public, it is invite-only, and is e2e encrypted. This way, the poster is still liable, and the service provider still makes a buck. Just can't use algorithms so the provider can't use virality. It's something similar to what I have done at 6groups.com
Really in the end that would push users to completely unmoderated forums, likely designed with protocols that are hard to block. The amount of bad actors and bad information would increase even further.
And, these small sites you're talking about are apt to be overly responsible because of the potential risk. "Oh no you said the word bullet" BANGO BONGO, you got the BAN YO.
But hey, most authoritarians would love it. They'll get a little bit of cash together and sue sites that talk about LGBT issues, or suicide prevention and say the content on those sites is terroristic in some way, further repressing speech.
Yep, as I said, internet would skew to a smaller web properties, and that's a good thing.
> push users to completely unmoderated forums
There's a counterexample -- Hacker News. Moderators here do a much better job at keeping it clean and focused on their intended audience than any of the big media. I'd be happy to see more of such focused forums than few huge "one-size-fits-all" platforms.
But they would not be able to do business in the US (financial transactions with advertisers/data brokers), that's right.
A lot of the complaints against 230 is that the sites are doing that too much, in ways that aren't to the liking of elected politicians, i.e. if twitter makes an anti-hate speech filter republicans say it's targeting conservatives, or if Facebook takes down an anti-fascism group democrats will say liberals are being targeted.
Forcing more legal liability on the major content hosting platforms won't make them go away, it will just make them into record labels or movie studios from the 1950s, where only an exclusive group of tightly controlled creatives get their mundane, uncontroversial content hosted. It's not going to create some cambrian explosion of alternative platforms - that already exists, and they don't get traffic because a one-stop shop for content is a natural monopoly.
We have YouTube, Twitch, Instagram, and TikTok. We have Facebook, Twitter, Reddit, and Hacker News. We have Netflix, Disney+, Peacock, and AppleTV.
Which of those is the “natural monopoly” for content?
I don’t know what that teaches us other than that capital, whether represented by algorithms or studio execs, tends to follow the culture as much as it creates it.