5 ms·

I was struck by a thought while reading the article- are the feeds of social media companies actually protected under section 230? My Facebook feed is more than

I was struck by a thought while reading the article- are the feeds of social media companies actually protected under section 230? My Facebook feed is more than just a simple serving of my friends content, it’s a work in and of itself (like a collection of essays). It seems like a stretch given that the law states that they can’t be considered the publisher of “information” provided by others, but possible as these sites are the ones engaged in picking content. The article mentions the roommates.com racial discrimination case. I wonder what a court would do if a site had some black box algorithm that ended up consistently showing a feed of rooms in one zip code to whites and rooms in another zip code to blacks, never overlapping. Edit- My point being- it might be possible to hold Facebook/Twitter/YouTube liable for the information in their feeds, particularly that which is surfaced out of order or from people a user isn’t subscribed to. But I’m not lawyer, nor do I play one on TV.

hemloc_io 6y ago
I've also wondered about this, I suspect because Facebook hired 75% of their lawyers in the last 3 months we will find out soon. I could see one argument about how an algorithm doesn't really editorialize. A metaphor that is often used is that social media companies is that they are more like a newsstand than an editorial board on a newspaper. A newsstand isn't under any obligation to sell all the papers available in their stand or place them in a specific order. I would expect the social media companies to say that their algorithms are closer to placing newspapers on a stand based on market conditions, not based on their personal opinions/eye. If that holds up to scrutiny we'll see.
- liaukovv 6y ago
  These websites are like the proverbial 10000 monkeys generating random symbols. Tech companies can form any narrative by picking parts they promote, does that still make them not responsible for the result?
- leereeves 6y ago
  Yes, in one sense an algorithm doesn't editorialize, not being intelligent, but humans can write an algorithm to editorialize for them. The editorial decisions are made while designing the algorithm. And if Facebook, knowing that outrage drives engagement, chooses to run an algorithm that displays the most engaging content, aren't they also choosing to display the most outrageous content?
  - laurent92 6y ago
    That is a good point, you not only demonstrated that choosing which comments to display is editorializing, but displaying the most outrageous content to provoke engagement is, in itself, an algorithm that incites violence.
lovecg 6y ago
That’s an interesting train of thought. I also wonder about content that’s been modified in some significant way (fact check labels not clearly separated from the content itself for example) by the provider: is there a point where the provider could become liable? I.e. if I made a joke website where you could post comments and it would auto-replace any mention of “topkai22” with “the notorious scoundrel topkai22”, could you sue me directly?
- zerocrates 6y ago
  Information that a provider creates itself isn't covered at all by 230, so they're potentially liable for things that don't originate from users like fact check labels no matter how they're applied. That's one of the many silly things about the usual 230 discourse: one of the earlier flashpoints was around things like Facebook and Twitter flagging "misinformation" and the like, when they don't actually get any protection under 230 for those labels, since they're writing them themselves.
  - topkai22 6y ago
    Have the "misinformation" flags been tested in court? It seems like that could possibly be covered by the "Good Samaritan" provision, although it seems it would hinge on if the definition of "restrict access to or availability of material" would including additional labeling rather than flat out blocking.
  - onionisafruit 6y ago
    Facebook went with the model of using third parties for fact checking. That lets them claim they are just a platform for the fact checkers.
- topkai22 6y ago
  I think that one is pretty clear- you are providing the phrase "notorious scoundrel", so you are the provider of that information. A subtler one is a site that only surfaced snippets about a subject. It wouldn't be that hard to write an algorithm that surfaces only negative segments about a subject, or to automatically quote out of context. Surely that is transformative, but the information is still coming from the original authors. Seems like that part might be untested law.
r_ms 6y ago
If a site has a black box algorithm that uses unlawful criteria (race) to limit the scope of searches, the site will not be immune under Section 230c and are liable if and only if they require the users to provide the information (race) as a condition of accessing its service in which case they act as “information content provider”
- froh 6y ago
  To my understanding fb would just need a number of your likes or shares to guess your gender, ethnicity, age, education level. You giving away your data this way, would that count as providing the information to control the information stream? And follow-up: This reminds me of the story of that data mining store chain which with high accuracy predicted pregnancy. A dad to a teenage daughter was very pissed when the chain started, in his perception, a personalized mail in coupon campaign to his teenage daughter "to become pregnant". When she actually already was. (The dad later apologized.) Nonetheless, the chain then concealed their knowledge, their targeted baby stuff and organic everything coupons, between just enough noise of tires, tools and men's sth sth, that it was no longer obvious to the innocent eye. Would that tactics be sufficient to white wash from section 230c obligations?
  - topkai22 6y ago
    Right- it's well known that discriminatory algorithms can evolve unintentionally. It's quite possible to accidentally create a feed that almost never shows black people white potential roommates and vice-versa. I don't think "the algorithm did it" would be necessarily be shield against liability. And, back to my orginal point, I think hosting content a person's profile page is one thing, and making the (probably automated) editorial decision to show 100,000 people that content in their streams is another.
    - BlueTemplar 6y ago
      Well, it's like in murder vs manslaughter, the intent matters (but not to the point to get you completely off the hook).
ricardo81 6y ago
>liable for Think 10-15 years ago looking at the concept of these centralised platforms it'd pretty obvious any kind of arbitration would not be scalable and therefore there wouldn't be. I mean, the business model depends on it. I'm not a fan of them being arbiters of truth, also not a fan of there being 10's of thousands of blatant copyright infractions on their platforms daily which they monetise. I don't think there's any other answer than decentralising groups on the web.
- topkai22 6y ago
  I think there needs to be a distinction between information seeking and passive interaction. If I go onto twitter and go to DPRK News (https://twitter.com/DPRK_News https://twitter.com/DPRK_News , quite a good parody account) it's clear that DPRK News is the publisher of all the information and twitter is the hoster. I'm seeking the information in the DPRK News feed. When I'm just flipping through twitter's infinite scroll then I'm not really seeking anything in particular than twitter is acting like a traditional editor- choosing the content I see. It's a different activity. I think platforms that choose content on behalf of the user need to have a different set of obligations than those that serve content based on user requests, an those obligations go up with scale. It's one thing to have a fake story or illegal content be seen by 100 friends, its another to feed it to 100,000 people.
  - ricardo81 6y ago
    Agreed. I'd also add it's important how you discover DPRK News in the first place e.g. if it was recommended somehow via the platform.
boomboomsubban 6y ago
Couldn't you make the same argument about stickied posts on old forums? I don't think how the content is displayed changes whether they are covered by section 230, and I think the anti-discriminaton exceptions would need to be more explicit to matter.
- BlueTemplar 6y ago
  It looks like to me that you can. When does a moderation choice becomes an editorial choice? Shouldn't there be some difference between the kind of moderation that removes content and the kind of moderation that promotes content?
  - boomboomsubban 6y ago
    >When does a moderation choice becomes an editorial choice? For the purpose of this law, never. Promoting a post does not make you the creator of that post.
    - dd36 6y ago
      It makes you responsible for what your user sees and doesn’t see. If you’re not presenting info transparently by time or popularity, you’re editorializing. Perhaps the reason there hasn’t been a big, precedent setting court case yet is the difficulty getting the data or seeing the editorializing. It seems like that veil is slowly coming down though through some of these court cases. What comment on a local news article does Facebook decide to show me to drive engagement? Why was it picked instead of another? I suppose one could get timestamps on their friends profiles and compare that to what was displayed to them. The idea that Facebook can make these editorial decisions at the coding level yet not be liable like a newspaper, is one thing that needs to change.
      - boomboomsubban 6y ago
        First, your opinion on how things should work is separate from the discussion I was taking part in. "Editorializing" isn't relevant to the protection section 230 provides. On your point, I'm not entirely sure what you want. Do you think Facebook should be sble face libel charges for how they present somebody else's post? Or do you want their ranking algorithm should be transparent. Should Google search results be considered the sane way?

hemloc_io 6y ago

I've also wondered about this, I suspect because Facebook hired 75% of their lawyers in the last 3 months we will find out soon. I could see one argument about how an algorithm doesn't really editorialize. A metaphor that is often used is that social media companies is that they are more like a newsstand than an editorial board on a newspaper. A newsstand isn't under any obligation to sell all the papers available in their stand or place them in a specific order. I would expect the social media companies to say that their algorithms are closer to placing newspapers on a stand based on market conditions, not based on their personal opinions/eye. If that holds up to scrutiny we'll see.

liaukovv 6y ago
These websites are like the proverbial 10000 monkeys generating random symbols. Tech companies can form any narrative by picking parts they promote, does that still make them not responsible for the result?
leereeves 6y ago
Yes, in one sense an algorithm doesn't editorialize, not being intelligent, but humans can write an algorithm to editorialize for them. The editorial decisions are made while designing the algorithm. And if Facebook, knowing that outrage drives engagement, chooses to run an algorithm that displays the most engaging content, aren't they also choosing to display the most outrageous content?
- laurent92 6y ago
  That is a good point, you not only demonstrated that choosing which comments to display is editorializing, but displaying the most outrageous content to provoke engagement is, in itself, an algorithm that incites violence.

lovecg 6y ago

That’s an interesting train of thought. I also wonder about content that’s been modified in some significant way (fact check labels not clearly separated from the content itself for example) by the provider: is there a point where the provider could become liable? I.e. if I made a joke website where you could post comments and it would auto-replace any mention of “topkai22” with “the notorious scoundrel topkai22”, could you sue me directly?

zerocrates 6y ago
Information that a provider creates itself isn't covered at all by 230, so they're potentially liable for things that don't originate from users like fact check labels no matter how they're applied. That's one of the many silly things about the usual 230 discourse: one of the earlier flashpoints was around things like Facebook and Twitter flagging "misinformation" and the like, when they don't actually get any protection under 230 for those labels, since they're writing them themselves.
- topkai22 6y ago
  Have the "misinformation" flags been tested in court? It seems like that could possibly be covered by the "Good Samaritan" provision, although it seems it would hinge on if the definition of "restrict access to or availability of material" would including additional labeling rather than flat out blocking.
- onionisafruit 6y ago
  Facebook went with the model of using third parties for fact checking. That lets them claim they are just a platform for the fact checkers.
topkai22 6y ago
I think that one is pretty clear- you are providing the phrase "notorious scoundrel", so you are the provider of that information. A subtler one is a site that only surfaced snippets about a subject. It wouldn't be that hard to write an algorithm that surfaces only negative segments about a subject, or to automatically quote out of context. Surely that is transformative, but the information is still coming from the original authors. Seems like that part might be untested law.

r_ms 6y ago

If a site has a black box algorithm that uses unlawful criteria (race) to limit the scope of searches, the site will not be immune under Section 230c and are liable if and only if they require the users to provide the information (race) as a condition of accessing its service in which case they act as “information content provider”

froh 6y ago
To my understanding fb would just need a number of your likes or shares to guess your gender, ethnicity, age, education level. You giving away your data this way, would that count as providing the information to control the information stream? And follow-up: This reminds me of the story of that data mining store chain which with high accuracy predicted pregnancy. A dad to a teenage daughter was very pissed when the chain started, in his perception, a personalized mail in coupon campaign to his teenage daughter "to become pregnant". When she actually already was. (The dad later apologized.) Nonetheless, the chain then concealed their knowledge, their targeted baby stuff and organic everything coupons, between just enough noise of tires, tools and men's sth sth, that it was no longer obvious to the innocent eye. Would that tactics be sufficient to white wash from section 230c obligations?
- topkai22 6y ago
  Right- it's well known that discriminatory algorithms can evolve unintentionally. It's quite possible to accidentally create a feed that almost never shows black people white potential roommates and vice-versa. I don't think "the algorithm did it" would be necessarily be shield against liability. And, back to my orginal point, I think hosting content a person's profile page is one thing, and making the (probably automated) editorial decision to show 100,000 people that content in their streams is another.
  - BlueTemplar 6y ago
    Well, it's like in murder vs manslaughter, the intent matters (but not to the point to get you completely off the hook).

ricardo81 6y ago

>liable for Think 10-15 years ago looking at the concept of these centralised platforms it'd pretty obvious any kind of arbitration would not be scalable and therefore there wouldn't be. I mean, the business model depends on it. I'm not a fan of them being arbiters of truth, also not a fan of there being 10's of thousands of blatant copyright infractions on their platforms daily which they monetise. I don't think there's any other answer than decentralising groups on the web.

topkai22 6y ago
I think there needs to be a distinction between information seeking and passive interaction. If I go onto twitter and go to DPRK News (https://twitter.com/DPRK_News https://twitter.com/DPRK_News , quite a good parody account) it's clear that DPRK News is the publisher of all the information and twitter is the hoster. I'm seeking the information in the DPRK News feed. When I'm just flipping through twitter's infinite scroll then I'm not really seeking anything in particular than twitter is acting like a traditional editor- choosing the content I see. It's a different activity. I think platforms that choose content on behalf of the user need to have a different set of obligations than those that serve content based on user requests, an those obligations go up with scale. It's one thing to have a fake story or illegal content be seen by 100 friends, its another to feed it to 100,000 people.
- ricardo81 6y ago
  Agreed. I'd also add it's important how you discover DPRK News in the first place e.g. if it was recommended somehow via the platform.

boomboomsubban 6y ago

Couldn't you make the same argument about stickied posts on old forums? I don't think how the content is displayed changes whether they are covered by section 230, and I think the anti-discriminaton exceptions would need to be more explicit to matter.

BlueTemplar 6y ago
It looks like to me that you can. When does a moderation choice becomes an editorial choice? Shouldn't there be some difference between the kind of moderation that removes content and the kind of moderation that promotes content?
- boomboomsubban 6y ago
  >When does a moderation choice becomes an editorial choice? For the purpose of this law, never. Promoting a post does not make you the creator of that post.
  - dd36 6y ago
    It makes you responsible for what your user sees and doesn’t see. If you’re not presenting info transparently by time or popularity, you’re editorializing. Perhaps the reason there hasn’t been a big, precedent setting court case yet is the difficulty getting the data or seeing the editorializing. It seems like that veil is slowly coming down though through some of these court cases. What comment on a local news article does Facebook decide to show me to drive engagement? Why was it picked instead of another? I suppose one could get timestamps on their friends profiles and compare that to what was displayed to them. The idea that Facebook can make these editorial decisions at the coding level yet not be liable like a newspaper, is one thing that needs to change.
    - boomboomsubban 6y ago
      First, your opinion on how things should work is separate from the discussion I was taking part in. "Editorializing" isn't relevant to the protection section 230 provides. On your point, I'm not entirely sure what you want. Do you think Facebook should be sble face libel charges for how they present somebody else's post? Or do you want their ranking algorithm should be transparent. Should Google search results be considered the sane way?