I didn’t see from the article if they have a definition of “the algorithm” that means anything?
Like, I know this is preaching to the choir here of all places, but you literally can’t have a social media without “an algorithm”, but I think we all know that’s not what they mean. But what actually do they mean? Do they mean “showing things from people you haven’t followed or subscribed to”, do they mean “showing posts or videos with stats that make them think you’ll like it”? Do they mean any selection algorithm that’s personalized to your account rather than site wide? So you could still have “top of the last 12 hours”, but everyone would have the same top?
Because every platform has more content on it per day than any human could consume that day. Something somewhere must choose what to show you and what order it’s in, because it can’t show everything. So I’m curious how they’ll define what “the algorithm” is…
Sure, but there can be some downsides to chronological order. I mean, right off the bat we need to know chronological order of what, because the firehose is all chaff no wheat. So maybe it’s firehose of only people I’ve followed, which means I’ll never discover anyone new, but maybe that’s okay for my goals.
But it also suffers from a starvation problem where if I have a friend that posts once a day, and a few friends and/or public figures or whatever that post many times an hour, I’ll probably never ever see what my quiet friend has to say because it’ll be lost in the noise.
And I’m sure some of you could say that’s my fault for following noisy accounts, but just because I want to see their things sometimes it doesn’t mean I only want to see their things. There’s room for an algorithm to skew rare-but-personal things above noisy-public-things without it being evil.
And then with something like Lemmy that’s kinda nuts because there’s lots and lots of stuff posted, and probably most of it is poo, and so I find “top 12 hours” or something to be much more value for effort, but that’s not chronological anymore.
You uh, didn’t use metacrawler, Altavista, ask Jeeves, yahoo, or Google(back in the day) much did you?
The internet has been more content than a human can experience in a lifetime since it was brought to the masses, and probably even before the pentium was created and a oc3 line was the dream of every computer nerd and geek
I did actually! But each of those used an algorithm to decide which things to show people based on their query. They must, because they couldn’t show everything.
So since they must choose “an algorithm” in the computer science sense, but not “the algorithm” in the pop reporting sense, the question becomes “which algorithms are the algorithm”, or more specifically which features are the actual problem.
Is it statistics? Is it personalization? Is it advertiser influence? That’s the question that’s interesting to me.
Thank you for sving me the trouble of typing basically that. heh.
It’s like when they talk about how to define porn and the best definition is basically “you know it when you see it”. But you can’t write law like that (except they can and do and it causes problems).
Chosen as in “subscriptions” / “followed” / “friends”?
So if it picks from among accounts you’ve opted-in to then it’s not the bad algorithm? That’s not meant as a gotcha, it’s a choice one could make to draw the line, but I don’t know if it’s the one they had in mind. There’s still room there to pick which order it shows your selected things in, but maybe that still mitigates the harm.
Does TikTok have the concept of following or subscribing? I don’t even know…
Things like Lemmy are a little trickier because our model is different. I can subscribe to a community, but then the content in that community is posted by strangers, and then those posts have comments by strangers, all of which require some kind of algorithm to curate. So in this definition showing me strangers shitposts is “no algorithm” if I’m subscribed to the shitpost community, but it isn’t allowed to suggest a shitpost if I’m not in that community?
It’s a line that could be drawn, but it remains to be seen if that’s the line they have in their minds.
Lemmy sorts by new, top and comment count, all of which are easily checkable, and I imagine hot and active are too. These aren’t algorithms because their methods aren’t hidden.
Right, I know they do. But “easily checkable” wasn’t our criteria before, this is a different criteria.
Is the criteria just that any algorithm that is transparent is okay? If there was a recommendation system that used your data, but the way they used it was open source, would that be “not an algorithm”? Because that’s also an option, but it’s a different option from this thread’s previous proposal that it be followers only or whatever.
I’m not really advocating for any particular choice here, I’m just saying that if they’re going to make a law around picking which algorithms are okay and which aren’t, they are going to need to figure out where those lines are, and I think that’s going to be harder than the buzzword makes it seem.
I didn’t see from the article if they have a definition of “the algorithm” that means anything?
Like, I know this is preaching to the choir here of all places, but you literally can’t have a social media without “an algorithm”, but I think we all know that’s not what they mean. But what actually do they mean? Do they mean “showing things from people you haven’t followed or subscribed to”, do they mean “showing posts or videos with stats that make them think you’ll like it”? Do they mean any selection algorithm that’s personalized to your account rather than site wide? So you could still have “top of the last 12 hours”, but everyone would have the same top?
Because every platform has more content on it per day than any human could consume that day. Something somewhere must choose what to show you and what order it’s in, because it can’t show everything. So I’m curious how they’ll define what “the algorithm” is…
“The algorithm” is just chronological order.
Which is the best algorithm to use.
Sure, but there can be some downsides to chronological order. I mean, right off the bat we need to know chronological order of what, because the firehose is all chaff no wheat. So maybe it’s firehose of only people I’ve followed, which means I’ll never discover anyone new, but maybe that’s okay for my goals.
But it also suffers from a starvation problem where if I have a friend that posts once a day, and a few friends and/or public figures or whatever that post many times an hour, I’ll probably never ever see what my quiet friend has to say because it’ll be lost in the noise.
And I’m sure some of you could say that’s my fault for following noisy accounts, but just because I want to see their things sometimes it doesn’t mean I only want to see their things. There’s room for an algorithm to skew rare-but-personal things above noisy-public-things without it being evil.
And then with something like Lemmy that’s kinda nuts because there’s lots and lots of stuff posted, and probably most of it is poo, and so I find “top 12 hours” or something to be much more value for effort, but that’s not chronological anymore.
You uh, didn’t use metacrawler, Altavista, ask Jeeves, yahoo, or Google(back in the day) much did you?
The internet has been more content than a human can experience in a lifetime since it was brought to the masses, and probably even before the pentium was created and a oc3 line was the dream of every computer nerd and geek
I did actually! But each of those used an algorithm to decide which things to show people based on their query. They must, because they couldn’t show everything.
So since they must choose “an algorithm” in the computer science sense, but not “the algorithm” in the pop reporting sense, the question becomes “which algorithms are the algorithm”, or more specifically which features are the actual problem.
Is it statistics? Is it personalization? Is it advertiser influence? That’s the question that’s interesting to me.
Thank you for sving me the trouble of typing basically that. heh.
It’s like when they talk about how to define porn and the best definition is basically “you know it when you see it”. But you can’t write law like that (except they can and do and it causes problems).
The algorithm here means showing content to one that wasn’t directly chosen, e.g. SEO on Google.
Sense 4 (Wiktionary)
Chosen as in “subscriptions” / “followed” / “friends”?
So if it picks from among accounts you’ve opted-in to then it’s not the bad algorithm? That’s not meant as a gotcha, it’s a choice one could make to draw the line, but I don’t know if it’s the one they had in mind. There’s still room there to pick which order it shows your selected things in, but maybe that still mitigates the harm.
Does TikTok have the concept of following or subscribing? I don’t even know…
Things like Lemmy are a little trickier because our model is different. I can subscribe to a community, but then the content in that community is posted by strangers, and then those posts have comments by strangers, all of which require some kind of algorithm to curate. So in this definition showing me strangers shitposts is “no algorithm” if I’m subscribed to the shitpost community, but it isn’t allowed to suggest a shitpost if I’m not in that community?
It’s a line that could be drawn, but it remains to be seen if that’s the line they have in their minds.
Lemmy sorts by new, top and comment count, all of which are easily checkable, and I imagine hot and active are too. These aren’t algorithms because their methods aren’t hidden.
Right, I know they do. But “easily checkable” wasn’t our criteria before, this is a different criteria.
Is the criteria just that any algorithm that is transparent is okay? If there was a recommendation system that used your data, but the way they used it was open source, would that be “not an algorithm”? Because that’s also an option, but it’s a different option from this thread’s previous proposal that it be followers only or whatever.
I’m not really advocating for any particular choice here, I’m just saying that if they’re going to make a law around picking which algorithms are okay and which aren’t, they are going to need to figure out where those lines are, and I think that’s going to be harder than the buzzword makes it seem.