14 ms·

By growing as large as they have, and by building automated systems to amplify content to mass audiences, they have acquired that role. It is unfortunate that

by jon37 6y ago

By growing as large as they have, and by building automated systems to amplify content to mass audiences, they have acquired that role. It is unfortunate that their control over their responsibility is unilateral and undemocratic. But at their scale, if they chose not to try and assess the accuracy of information, but instead to blindly amplify it based on engagement metrics, that is also a political choice.

bobthepanda 6y ago
One possible option that never gets discussed is to nuke the amplification methods. If we stop recommending content automatically this ceases to be a problem.
- crazygringo 6y ago
  They are nuking them right now (temporarily). It's not just being discussed, it's being done. But most of the time people like getting content recommended. It's what consumers want.
  - munk-a 6y ago
    I disagree, this isn't the nuclear option, the nuclear option is forcing these platforms to have a more editorial role in the content they're serving and that comes with a whole bunch of good and a boatload of bad. Gigantic unmoderated platforms existing like this that promote random snippets of speech to drive user engagement and ad-revenue is a thing that shouldn't exist. The problem we still haven't solved is how to specifically kill off platforms of this type without killing forums and discussion boards in general. I think there is a distinction there but I'm not certain precisely what defines it - but if anyone figures it out please let us all know!
    - bobthepanda 6y ago
      we've had forums and discussion boards for decades now that do not have recommendation features. I don't see why we can't put that genie back in the bottle. IMO the moment you start highlighting things that people didn't explicitly ask for, it's an endorsement.
      - munk-a 6y ago
        I think it's like Gerrymandering - yea we can all tell when it's gotten to stupid levels but the supreme court wasn't wrong to want a definition of where the line between "okay" and "bonkers" is. I personally think the decision could've been a bit more aggressive against gerrymandering but we do need some clear line to say "If you're beyond this you're doing an illegal thing" - and while we could close in on that line over time with a slow accumulation of precedent it'd be a lot cleaner to have a decent measure.
  - xboxnolifes 6y ago
    > It's what consumers want. Do consumers want it, or is it merely taking advantage of some more subconscious human behavior patterns. And if the latter, is this something that is bad for humankind?
    - robbiep 6y ago
      Oh mate - definitely something that the dopamine dealers need to confront at some point. ‘Want’ vs actual material benefit to life
    - reducesuffering 6y ago
      Tobacco 2.0
  - bobthepanda 6y ago
    > But most of the time people like getting content recommended. It's what consumers want. Do they, or do they just boost some KPI that suits proxy for actual utility? Anecdotally even in non-tech circles most of my friends complain about how bad recommended content has gotten, or roll their eyes at whatever "personalized" ad for garbage they've been recommended.
  - jon37 6y ago
    Consumers want a lot of things with negative externalities - goods that cost less because they're produced with slave labor, transportation that emits greenhouse gases, etc. Their preference shouldn't trump the obligation not to harm third parties. Automated recommendations of a human-curated set of content - e.g. Netflix recommendations for its suite of programming - are much less objectionable, because they can't amplify anything the organization has not intentionally decided to present. It's the combination of UGC and ML recommendations that presents problems.
  - RIMR 6y ago
    Consumers also want cocaine, but that doesn't mean you get to sell it to them with impunity.
    - shadowgovt 6y ago
      We cross a threshold if we treat information like cocaine. A threshold that calls the First Amendment in the US into question.
      - cutemonster 6y ago
        Is lying and deceiving people fine, according to the 1st amendment? For example foreign states that pays armies of internet trolls to in effect choose president in the US -- is that what the 1st amendment wants to happen I think "information" can kill more people than cocaine, is more dangerous
        shadowgovt 6y ago
        Don't misunderstand that I'm a First Amendment absolutist. Indeed, I mean exactly what I said: the questions you raise call the First Amendment into question.
        cutemonster 6y ago
        Hmm not sure I understand. You're saying that maybe the 1st amendment could be edited a bit so that it was more clear that such things aren't okay? I'm thinking people interpret "free speech" in sometimes quite different ways
    - nobody9999 6y ago
      >Consumers also want cocaine, but that doesn't mean you get to sell it to them with impunity. The appropriate situation for those that want cocaine is something similar to the rules around purchasing/possessing/using alcohol. From an economic standpoint (increased tax revenue, reduced spending on "enforcement" and incarceration, increased economic output because fewer people are in prison, etc.) and a societal standpoint (more resources available to the 2-5% of folks who end up with dependency problems, reduced property crime, not harming communities with significant numbers of residents being pulled out of the community and incarcerated, etc.) As such, there's no good reason for any mind altering substances to be illegal. Rather, they should be regulated and taxed appropriately.
  - grishka 6y ago
    As someone who's worked at a big social media company — no, that's not at all what consumers want. They want chronological feeds with zero garbage mixed in. It's okay to have a separate recommendations feed, some (not all!) people want to discover new things, but it's totally not okay to meddle with the main one, and it's nothing but mockery to give users no control over it. People also want their preferences respected, they certainly don't want them reset every now and then. The only reason people keep using services like Twitter is because their network keeps them there.
    - crazygringo 6y ago
      Well I guess it depends on making a distinction between what consumers think they want, versus what they actually do. Yes, people say they don't want recommendations, because 95% of them are irrelevant. But then the 5% (or 2% or 0.5%) turn out to be super-relevant, and they find new people to follow that they love, and learn about things they love, and the experience in the end turns out to be a net positive. Their actions show that it's valuable in the end. Otherwise the feature wouldn't exist at all. Recommendations aren't advertisements, sites don't make money off them -- sites use them because people genuinely find things that lead them to use the sites more.
      - grishka 6y ago
        I'm not denying the undeniable fact that some people sometimes want to discover new things. I'm just saying that it's absolutely possible to have it done in a respectful manner. No one, ever, under any circumstances, likes or wants to be manipulated, be it overtly or by having their subconscious played with — period. Adding non-configurable extra anything into people's newsfeeds, be it recommended posts, ads, or "people you may know" blocks, is a crime against user-frendliness. Those who do want to discover new things, will simply open the "discover"/"explore" tab that contains a dedicated recommended content feed on their own. There is no need to nudge anyone to anything. People aren't stupid if you don't build your UI/UX around the assumption that they are. They also like transparent, understandable algorithms. Chronological feed of (only) the people one follows is as transparent as it gets. A chronological feed with some recommendations mixed in is more opaque and confusing. An algorithmic feed is an epitome of opaqueness. Opaqueness naturally drives users away because it doesn't exactly instill confidence that their posts will reach their followers. Another example: do you understand what the "see less often" button in Twitter does? No one does. No one likes cryptic algorithmic bullshit forced on them with no way to disable it. Choice is very important. > versus what they actually do. Do manipulations work? Of course they do. Are people happy when they are manipulated? Of course they are not. > Recommendations aren't advertisements, sites don't make money off them They absolutely do. Recommendations aren't there because Twitter wants to be helpful — they'd be more user-respecting as I said above if that was the case. They're there because they drive engagement metrics up, and those in turn translate into someone's KPI.
Nuzzerino 6y ago
Then they shouldn't have become that large to begin with.
konjin 6y ago
>But at their scale, if they chose not to try and assess the accuracy of information, but instead to blindly amplify it based on engagement metrics, that is also a political choice. No that's an apolitical choice. The political choice was not doing this in 2012 when it was Obama benefiting from it.
hda2 6y ago
if a social network grows beyond a certain threshold, say 100M, it should be forced by law to behave like a democracy.
mkolodny 6y ago
I also wish that these decisions were more democratic. At the same time, personally, I think that the folks who made these changes did a great job. They're helping preserve American democracy. In particular, I appreciate: "10/2019 - Banned all political ads on Twitter, including ads from state-controlled media" I hope Facebook employees are taking notes. I'm also a huge fan of: "...we will label Tweets that falsely claim a win for any candidate and will remove Tweets that encourage violence or call for people to interfere with election results or the smooth operation of polling places." Does anyone know whether the next one is official US policy, or whether it's just Twitter's policy? "To determine the results of an election in the US, we require either an announcement from state election officials, or a public projection from at least two authoritative, national news outlets that make independent election calls." Mail-in ballots can arrive at the voting office as late as November 20th this year (depending on your state.. you still have to mail your ballot out by November 3rd, though) [0]. With so many people voting by mail this year, we might not know the election results until November 20th. I hope that election officials (and Twitter officials) will take that into account. [0] https://www.sos.ca.gov/elections/voter-registration/vote-mail/#vote-by-mail https://www.sos.ca.gov/elections/voter-registration/vote-mai...

by jon37 6y ago

bobthepanda 6y ago

One possible option that never gets discussed is to nuke the amplification methods. If we stop recommending content automatically this ceases to be a problem.

crazygringo 6y ago
They are nuking them right now (temporarily). It's not just being discussed, it's being done. But most of the time people like getting content recommended. It's what consumers want.
- munk-a 6y ago
  I disagree, this isn't the nuclear option, the nuclear option is forcing these platforms to have a more editorial role in the content they're serving and that comes with a whole bunch of good and a boatload of bad. Gigantic unmoderated platforms existing like this that promote random snippets of speech to drive user engagement and ad-revenue is a thing that shouldn't exist. The problem we still haven't solved is how to specifically kill off platforms of this type without killing forums and discussion boards in general. I think there is a distinction there but I'm not certain precisely what defines it - but if anyone figures it out please let us all know!
  - bobthepanda 6y ago
    we've had forums and discussion boards for decades now that do not have recommendation features. I don't see why we can't put that genie back in the bottle. IMO the moment you start highlighting things that people didn't explicitly ask for, it's an endorsement.
    - munk-a 6y ago
      I think it's like Gerrymandering - yea we can all tell when it's gotten to stupid levels but the supreme court wasn't wrong to want a definition of where the line between "okay" and "bonkers" is. I personally think the decision could've been a bit more aggressive against gerrymandering but we do need some clear line to say "If you're beyond this you're doing an illegal thing" - and while we could close in on that line over time with a slow accumulation of precedent it'd be a lot cleaner to have a decent measure.
- xboxnolifes 6y ago
  > It's what consumers want. Do consumers want it, or is it merely taking advantage of some more subconscious human behavior patterns. And if the latter, is this something that is bad for humankind?
  - robbiep 6y ago
    Oh mate - definitely something that the dopamine dealers need to confront at some point. ‘Want’ vs actual material benefit to life
  - reducesuffering 6y ago
    Tobacco 2.0
- bobthepanda 6y ago
  > But most of the time people like getting content recommended. It's what consumers want. Do they, or do they just boost some KPI that suits proxy for actual utility? Anecdotally even in non-tech circles most of my friends complain about how bad recommended content has gotten, or roll their eyes at whatever "personalized" ad for garbage they've been recommended.
- jon37 6y ago
  Consumers want a lot of things with negative externalities - goods that cost less because they're produced with slave labor, transportation that emits greenhouse gases, etc. Their preference shouldn't trump the obligation not to harm third parties. Automated recommendations of a human-curated set of content - e.g. Netflix recommendations for its suite of programming - are much less objectionable, because they can't amplify anything the organization has not intentionally decided to present. It's the combination of UGC and ML recommendations that presents problems.
- RIMR 6y ago
  Consumers also want cocaine, but that doesn't mean you get to sell it to them with impunity.
  - shadowgovt 6y ago
    We cross a threshold if we treat information like cocaine. A threshold that calls the First Amendment in the US into question.
    - cutemonster 6y ago
      Is lying and deceiving people fine, according to the 1st amendment? For example foreign states that pays armies of internet trolls to in effect choose president in the US -- is that what the 1st amendment wants to happen I think "information" can kill more people than cocaine, is more dangerous
      - shadowgovt 6y ago
        Don't misunderstand that I'm a First Amendment absolutist. Indeed, I mean exactly what I said: the questions you raise call the First Amendment into question.
        cutemonster 6y ago
        Hmm not sure I understand. You're saying that maybe the 1st amendment could be edited a bit so that it was more clear that such things aren't okay? I'm thinking people interpret "free speech" in sometimes quite different ways
  - nobody9999 6y ago
    >Consumers also want cocaine, but that doesn't mean you get to sell it to them with impunity. The appropriate situation for those that want cocaine is something similar to the rules around purchasing/possessing/using alcohol. From an economic standpoint (increased tax revenue, reduced spending on "enforcement" and incarceration, increased economic output because fewer people are in prison, etc.) and a societal standpoint (more resources available to the 2-5% of folks who end up with dependency problems, reduced property crime, not harming communities with significant numbers of residents being pulled out of the community and incarcerated, etc.) As such, there's no good reason for any mind altering substances to be illegal. Rather, they should be regulated and taxed appropriately.
- grishka 6y ago
  As someone who's worked at a big social media company — no, that's not at all what consumers want. They want chronological feeds with zero garbage mixed in. It's okay to have a separate recommendations feed, some (not all!) people want to discover new things, but it's totally not okay to meddle with the main one, and it's nothing but mockery to give users no control over it. People also want their preferences respected, they certainly don't want them reset every now and then. The only reason people keep using services like Twitter is because their network keeps them there.
  - crazygringo 6y ago
    Well I guess it depends on making a distinction between what consumers think they want, versus what they actually do. Yes, people say they don't want recommendations, because 95% of them are irrelevant. But then the 5% (or 2% or 0.5%) turn out to be super-relevant, and they find new people to follow that they love, and learn about things they love, and the experience in the end turns out to be a net positive. Their actions show that it's valuable in the end. Otherwise the feature wouldn't exist at all. Recommendations aren't advertisements, sites don't make money off them -- sites use them because people genuinely find things that lead them to use the sites more.
    - grishka 6y ago
      I'm not denying the undeniable fact that some people sometimes want to discover new things. I'm just saying that it's absolutely possible to have it done in a respectful manner. No one, ever, under any circumstances, likes or wants to be manipulated, be it overtly or by having their subconscious played with — period. Adding non-configurable extra anything into people's newsfeeds, be it recommended posts, ads, or "people you may know" blocks, is a crime against user-frendliness. Those who do want to discover new things, will simply open the "discover"/"explore" tab that contains a dedicated recommended content feed on their own. There is no need to nudge anyone to anything. People aren't stupid if you don't build your UI/UX around the assumption that they are. They also like transparent, understandable algorithms. Chronological feed of (only) the people one follows is as transparent as it gets. A chronological feed with some recommendations mixed in is more opaque and confusing. An algorithmic feed is an epitome of opaqueness. Opaqueness naturally drives users away because it doesn't exactly instill confidence that their posts will reach their followers. Another example: do you understand what the "see less often" button in Twitter does? No one does. No one likes cryptic algorithmic bullshit forced on them with no way to disable it. Choice is very important. > versus what they actually do. Do manipulations work? Of course they do. Are people happy when they are manipulated? Of course they are not. > Recommendations aren't advertisements, sites don't make money off them They absolutely do. Recommendations aren't there because Twitter wants to be helpful — they'd be more user-respecting as I said above if that was the case. They're there because they drive engagement metrics up, and those in turn translate into someone's KPI.

Nuzzerino 6y ago

Then they shouldn't have become that large to begin with.

konjin 6y ago

>But at their scale, if they chose not to try and assess the accuracy of information, but instead to blindly amplify it based on engagement metrics, that is also a political choice. No that's an apolitical choice. The political choice was not doing this in 2012 when it was Obama benefiting from it.

hda2 6y ago

if a social network grows beyond a certain threshold, say 100M, it should be forced by law to behave like a democracy.

mkolodny 6y ago

I also wish that these decisions were more democratic. At the same time, personally, I think that the folks who made these changes did a great job. They're helping preserve American democracy. In particular, I appreciate: "10/2019 - Banned all political ads on Twitter, including ads from state-controlled media" I hope Facebook employees are taking notes. I'm also a huge fan of: "...we will label Tweets that falsely claim a win for any candidate and will remove Tweets that encourage violence or call for people to interfere with election results or the smooth operation of polling places." Does anyone know whether the next one is official US policy, or whether it's just Twitter's policy? "To determine the results of an election in the US, we require either an announcement from state election officials, or a public projection from at least two authoritative, national news outlets that make independent election calls." Mail-in ballots can arrive at the voting office as late as November 20th this year (depending on your state.. you still have to mail your ballot out by November 3rd, though) [0]. With so many people voting by mail this year, we might not know the election results until November 20th. I hope that election officials (and Twitter officials) will take that into account. [0] https://www.sos.ca.gov/elections/voter-registration/vote-mail/#vote-by-mail https://www.sos.ca.gov/elections/voter-registration/vote-mai...