The evidence shows an abrupt change in engagement, but that doesn't necessarily mean the algorithm was changed. The title is misleading as the study didn't directly conclude algorithm changes were the cause.
Interesting observation. From my experience, any abrupt shift in a metric is normally due to external factors rather than internal ones. Something that academics seem to miss entirely.
Indeed. As others have noted, that would be when the attempted assassination of Trump was, shortly followed by Elon endorsing him and then forming his PAC [1]. I'd expect such activities to raise interest in his activities and statements, so this entire thing seems to be a staggering miss at asking very basic questions before leaping to conclusions.
I think the left underestimates what a shot of adrenaline the assassination attempt injected into the right. It may have swung the election all by itself. It got people off of the couch.
> a structural break for Musk's metrics around July 13, 2024" following which his view counts increased by 138.27% and retweets increased by 237.94%, with a similarly large increase for favourites
I can't judge how legitimate this particular report is. From personal experience with an account I haven't used in double digit years and deleted recently, my feed was exclusively Musk tweets without it being my choice. So anecdotally I have every reason to believe that Musk inflates his own... everything, probably.
Using your company to artificially inflate your own importance somehow strikes me as extremely pathetic. It's like getting an employee to stuff your crotch to make things look (but not be!) bigger.
> Musk made his endorsement of Trump on July 13, 2024. The analysis suggests that Musk did more than merely give verbal support for Trump, but also changed X's algorithm on the same date to systematically promote his own and other prominent pro-Republican acounts
July 13 was the day of the assassination attempt...
The problem is that it takes a level of personal growth to have the wherewithall to understand the truth of such matters of truth and morality.
Put simply, it takes a morally dilligent, selfless, compassionate person to clearly judge moral situations. Dunning-Kruger is as applicable to personal morality as it is to technical excellence.
By coincidence I made a new temporarily account on Twitter ( new phone and picked the wrong email, didn't knew what combo it was with username and email).
The amount of bullshit that cropped up my feed immediately was nuts ( calling it right wing might be fitting).
I logged onto my old account afterwards and everything was normal again.
The open-sourcing of the X's feed algorithm is a typical Musk's smokescreen.
First, while the code was open sourced, the configs and data that determines its behavior wasn't. Any researcher that has some knowledge of the matter will tell you this.
Second, that code was last updated in July 2023 [1]. News feed algorithms are updated regularly, especially around the time of US elections [2].
Sure it's open source but that doesn't mean the public repo is in any way up to date. Especially given that the repos only have a handful of commits from the few months after the repos were published and they haven't been updated in well over a year.
How is this defined and why does everything suddenly qualify.
Don't get me wrong I don't like singular influence like this either, but what's the legality and how do we draw the like between this and any other attempts to sway public opinion by tuning content?
I think the easiest way to read these types of comments is not that they're actually election interference or even really accusing the behavior of being election interference but making a joke about all the nonsense that was called election interference in 2020 and pointing out the hypocrisy of not caring about the same behavior now that they're engaging in it.
How exactly is "election interference" defined? Because it seems to extend to any action that effects the outcome of an election. Is campaigning election interference? What about voting?
If the lottery will not have repercussions, or if it would have had repercussions if the other side won, I don't see how the checks and balances can be said to be functional.
It probably was interference, however note that removing left bias has exactly the same result as adding right bias.
I guess what I am trying to say is measuring bias sounds like a long and profitable academic career in itself. That is, hard to define, vague methodology, and uncertain results.
I thought the conversation was interesting. This post is clearly related to the interests of HN and it was going for the front page. Which specific guidelines did it break, if any?
This submission was about an algorithm of X, which clearly is on topic, particularly nowadays when millions of users migrate to Bluesky.
The study doesn't offer any conclusion, but it analyses some data and offers a potential interpretation. Other interpretations can be provided and discussed. The post was heading for the front page. Personally, I was very interested in the discussion.
The study is speculation about supposed changes to the algorithm based off of subsequent changes in trends in sheer ignorance to other events that transpired on the dates specified.
Basically arguing that correlation is causation and placing the cart before the horse. This isn't good grounds for insightful or even useful conversations and this was already breaking the guidelines by being a political subject matter about celebrities.
First, you don't understand the methodology of this study. Difference in differences takes into account external events by assuming that two different time series are affected by external events in the same way. In this case, the two time series correspond to the number of views of Republican-leaning and Democrat-leaning users on X. Thus, the methodology takes into account external events.
Second, the speculation you're writing about is an interpretation of the new data analysis.
Third, the authors don't argue that correlation is causation. If you still think they do, then please quote the sentence where they do that.
Honestly, it looks like this thread is being taken down because some people don't like it, including you, which goes against HN guidelines.
You're changing your arguments whenever I address the previous one. Have you noticed?
> Hacker News biases heavily against Trump and Musk
I imagine you're writing this only because you are biased for Musk (and/or Trump). Many people in Hacker News seek truth. Clearly, you're not the one, and you're deciding for this community that it shouldn't seek it either.
>You're changing your arguments whenever I address the previous one. Have you noticed?
Nope, I already pointed out the study is speculative bullshit: The axiom is "potential algorithmic bias around July 13, 2024".
"Potential" bias (read: changes) around one of the most significant days in the entire election cycle is why X suddenly shifted further Right from July 13 onwards?
So much happened on July 13 that changes to the algorithm (if any) at most played a minor factor, let alone "potential" changes that the study doesn't even have the gall to affirmatively declare.
Sincerely: Stop wasting my (our?) time with this bullshit.
>I imagine you're writing this only because you are biased for Musk (and/or Trump).
You are correct that I like Trump and Musk, I voted for Trump and I continue to be amazed at the crazy stuff Musk pulls off. A look through my comment history would indicate that, too.
>Many people in Hacker News seek truth.
If by "truth" you mean primarily Left/Progressive "truth". I will break the guidelines by saying this, but the audience here really isn't that different from Reddit with regards to politics.
This is to be expected, of course: Hacker News is a forum for tech enthusiasts/professionals and small cap investors operated by venture capitalists, and most if not all of those demographics are heavily Left/Progressive.
One thing I will point out that I think is fairly unique to Hacker News is the sheer level of resentment towards Musk. You can't even have a conversation about the guy unless it's to shit on him. If most other communities suffer Trump Derangement Syndrome, Hacker News suffers Musk Derangement Syndrome.
By truth I mean scientific discourse centered on evidence and freedom of speech. Unfortunately, Musk's and Trump's "freedom of speech" includes freedom to disinform the public to achieve their goals, so it isn't equivalent to everyone's freedom of speech. By taking down this post, you're acting against everyone's freedom of speech.
The study is honest, because it doesn't make any conclusions, just provides new data analysis, potential interpretations, and a call for further research. You haven't responded to my arguments, e.g., about how the study takes into account external events.
It sounds like you're misperceiving the post because of your personal biases, but at least you're upfront about it, which I appreciate. You're taking down the post, because you don't like it. That's against Hacker News guidelines. This kind of behavior needs more attention, just like the behaviors of Musk and Trump. We need more transparency, not less, so I appreciate that you're upfront.
BTW, I haven't voted for Harris nor Trump. My perfect (non-existing) candidate would take the best from each of them. While I appreciate various things about Trump and Musk, I'm critical about the fact that they both disinform the public to achieve their personal goals. That's unethical.
>By taking down this post, you're acting against everyone's freedom of speech.
As far as I understand flagging on Hacker News, everyone's free to flag anything and becomes [flagged] upon receiving enough such flags. So it's not just me or the guy who posted about flagging, enough people said this thread wasn't worth Hacker News's time.
>The study is honest, because it doesn't make any conclusions,
That makes it worse because that means I spend my time to achieve nothing. I don't need a rocket scientist to tell me something obvious happened for speculatory allegations.
>just provides new data analysis,
Musk has been quite clear about siding with the Right/Conservatives/Liberals/Republicans when it comes to his politics, and X being his personal property is very well known to lean that way at this point.
>potential interpretations, a call for further research.
A study of knowns like this one should put forth a declaration in the absolute.
>how the study takes into account external events.
The study alleges a potential change in algorithm based off certain posts' performance metrics and Musk's endorsement of Trump. Essentially arguing correlation is causation and without even declaring there was causation to merit the study. That is bullshit, I ask the study and those who disseminate it to please stop wasting my time.
>You're taking down the post, because you don't like it.
I flagged this thread because it is not conducive to thoughtful and insightful debate. The study refuses to declare any conclusions or accusations, just potentials and maybes, and the logic is flawed from the outset because it strives to draw a desired conclusion ("there was (maybe) algorithm changes") first and then the steps to get there at any cost.
Even aside that, this thread is also about politics surrounding celebrities which are by themselves also against guidelines.
This thread and study could have just as likely been about Zuckerberg and Facebook or Spez and Reddit or dang and Hacker News, I would have still flagged it all the same.
>I'm critical about the fact that they both disinform the public to achieve their personal goals. That's unethical.
It's also unethical to publish a study that throws out allegations without merit for the express purpose of driving narratives, like this one.
Can the study cite how, where, and when X changed its algorithm? If not, using the term "potentially" to avoid outright libel is cowardly; anyone lodging accusations should have the decency to own it. Does the logic in the study make sense? Of course not, correlation is not causation is one of the first rules of logic most of us should learn and respect.
That's why I flagged this thread and study. It's worthless, it's flawed, and it's unethical and cowardly.
You continue falling to the straw man fallacy by bringing over and over the correlation is causation criticism, but the study doesn't make causal claims...
I understand that an analysis like that is not conducive to thoughtful and insightful debate for closed-minded people who do not care, or even oppose, to learn things that don't align with their interests.
You continue to ignore my responses to your points, e.g., about the fact that the study takes care of external events by applying difference-in-differences methodology, and continue writing accusations without being specific.
If anything, disinformation and censorship of research are worthless, flawed, unethical, and cowardly.
This is interesting, but let's not have a takeaway that the problem is isolated to bad actors purchasing social networking sites. Even on supposedly neutral platforms like YouTube or Facebook you'll see bad actors successfully manipulate the algorithmic process to help shape opinion.
I find it interesting to reflect on the fact that -- despite being raised on a diet of science fiction scenarios about dystopias -- my generation built these machines. In hindsight, it's blindingly obvious that algorithms would be manipulated by the wealthy and powerful to service their agendas. Of _course_ they would, every incentive was plainly there from the start. But we figured it wouldn't be a problem because it was us who built them, and we weren't the baddies.
I say this not to point the finger at other people. I worked on a major social media system -- and I was one of those who defended the system as it was being created. And we all should have known better.
The evidence shows an abrupt change in engagement, but that doesn't necessarily mean the algorithm was changed. The title is misleading as the study didn't directly conclude algorithm changes were the cause.
I doubt anyone is alleging Elon Musk has control over Google's search. Google Trends suggests the same kind of interest change at exactly the same point in mid-July: https://trends.google.com/trends/explore?date=2024-01-01%202...
Interesting observation. From my experience, any abrupt shift in a metric is normally due to external factors rather than internal ones. Something that academics seem to miss entirely.
Indeed. As others have noted, that would be when the attempted assassination of Trump was, shortly followed by Elon endorsing him and then forming his PAC [1]. I'd expect such activities to raise interest in his activities and statements, so this entire thing seems to be a staggering miss at asking very basic questions before leaping to conclusions.
[1]: https://en.wikipedia.org/wiki/America_PAC
Yeah like an assassination attempt.
Twitter algorithm changes can presumably play some role in driving what people Google for.
Presumably they can lead to assassination attempts, too. But it becomes hard to prove.
This could be the result of the news feed change at X and Musk's endorsement of Trump...
For the unaware or the forgetful, July 13, 2024 was the day someone tried to assassinate Trump.[1]
I am not surprised that Trump and Republicans immediately garnered more attention as a consequence, rightfully or otherwise.
[1]: https://en.wikipedia.org/wiki/Attempted_assassination_of_Don...
The obvious answer is that Trumps assassination attempt were due to the twitter algorithm change that happened that day!!!
I think the left underestimates what a shot of adrenaline the assassination attempt injected into the right. It may have swung the election all by itself. It got people off of the couch.
Trump paid the guy to do it for this exact reason
Premium accounts get a boost.
You'd struggle to convince me the number of Premium registrations wasn't heavily biased to the Right.
> a structural break for Musk's metrics around July 13, 2024" following which his view counts increased by 138.27% and retweets increased by 237.94%, with a similarly large increase for favourites
I can't judge how legitimate this particular report is. From personal experience with an account I haven't used in double digit years and deleted recently, my feed was exclusively Musk tweets without it being my choice. So anecdotally I have every reason to believe that Musk inflates his own... everything, probably.
Using your company to artificially inflate your own importance somehow strikes me as extremely pathetic. It's like getting an employee to stuff your crotch to make things look (but not be!) bigger.
[flagged]
Earlier: https://news.ycombinator.com/item?id=42125818
> Musk made his endorsement of Trump on July 13, 2024. The analysis suggests that Musk did more than merely give verbal support for Trump, but also changed X's algorithm on the same date to systematically promote his own and other prominent pro-Republican acounts
July 13 was the day of the assassination attempt...
"Science"
[dead]
[flagged]
Correct. Nothing about Musk a person indicates that he cares about truth or justice or anything except his own personal wealth and power.
The problem is that it takes a level of personal growth to have the wherewithall to understand the truth of such matters of truth and morality.
Put simply, it takes a morally dilligent, selfless, compassionate person to clearly judge moral situations. Dunning-Kruger is as applicable to personal morality as it is to technical excellence.
[flagged]
By coincidence I made a new temporarily account on Twitter ( new phone and picked the wrong email, didn't knew what combo it was with username and email).
The amount of bullshit that cropped up my feed immediately was nuts ( calling it right wing might be fitting).
I logged onto my old account afterwards and everything was normal again.
Just my personal experience with it
Musk has said many times that the algorithm is open source. This guy is suggesting an algorithm change but not actually showing one.
https://github.com/twitter/the-algorithm - Last commit: July 13th, 2023
https://github.com/twitter/the-algorithm-ml - Last commit: April 6th, 2023
Either the algorithm hasn't been updated for over a year (unlikely) or the public version is abandoned.
The open-sourcing of the X's feed algorithm is a typical Musk's smokescreen.
First, while the code was open sourced, the configs and data that determines its behavior wasn't. Any researcher that has some knowledge of the matter will tell you this.
Second, that code was last updated in July 2023 [1]. News feed algorithms are updated regularly, especially around the time of US elections [2].
[1] https://github.com/twitter/the-algorithm
[2] https://www.science.org/doi/10.1126/science.adt2983
Sure it's open source but that doesn't mean the public repo is in any way up to date. Especially given that the repos only have a handful of commits from the few months after the repos were published and they haven't been updated in well over a year.
https://github.com/twitter/the-algorithm
https://github.com/twitter/the-algorithm-ml
wow, "view count findings suggest possible recommendation algorithm bias", much science
So, election interference?
How is this defined and why does everything suddenly qualify.
Don't get me wrong I don't like singular influence like this either, but what's the legality and how do we draw the like between this and any other attempts to sway public opinion by tuning content?
My understanding of US election law is that all of this is legal (in case this really did take place). If it's moral is a different question.
I think the easiest way to read these types of comments is not that they're actually election interference or even really accusing the behavior of being election interference but making a joke about all the nonsense that was called election interference in 2020 and pointing out the hypocrisy of not caring about the same behavior now that they're engaging in it.
This is the most polite "whoosh" I've ever had levied against me. Thanks!
How exactly is "election interference" defined? Because it seems to extend to any action that effects the outcome of an election. Is campaigning election interference? What about voting?
Which will have nothing done about it, just like his "lottery" he was running to promote trump.
If the lottery will not have repercussions, or if it would have had repercussions if the other side won, I don't see how the checks and balances can be said to be functional.
It was only election interference when Twitter’s former owners did it.
It probably was interference, however note that removing left bias has exactly the same result as adding right bias.
I guess what I am trying to say is measuring bias sounds like a long and profitable academic career in itself. That is, hard to define, vague methodology, and uncertain results.
Flagging this because a toot isn't news and the discussion isn't discussing the paper, everybodys just going off vibes in here.
I thought the conversation was interesting. This post is clearly related to the interests of HN and it was going for the front page. Which specific guidelines did it break, if any?
>Off-Topic: Most stories about politics, or crime, or sports, or celebrities, unless they're evidence of some interesting new phenomenon.
This was further exacerbated because the linked study seemingly ignores what happened on the day specified to demonstrate their narrative.
Essentially, the linked study placed the cart before the horse.
This submission was about an algorithm of X, which clearly is on topic, particularly nowadays when millions of users migrate to Bluesky.
The study doesn't offer any conclusion, but it analyses some data and offers a potential interpretation. Other interpretations can be provided and discussed. The post was heading for the front page. Personally, I was very interested in the discussion.
Was any HN guideline broken?
The most important guideline of all; thou shalt not make weird billionaires look bad.
The study is speculation about supposed changes to the algorithm based off of subsequent changes in trends in sheer ignorance to other events that transpired on the dates specified.
Basically arguing that correlation is causation and placing the cart before the horse. This isn't good grounds for insightful or even useful conversations and this was already breaking the guidelines by being a political subject matter about celebrities.
First, you don't understand the methodology of this study. Difference in differences takes into account external events by assuming that two different time series are affected by external events in the same way. In this case, the two time series correspond to the number of views of Republican-leaning and Democrat-leaning users on X. Thus, the methodology takes into account external events.
Second, the speculation you're writing about is an interpretation of the new data analysis.
Third, the authors don't argue that correlation is causation. If you still think they do, then please quote the sentence where they do that.
Honestly, it looks like this thread is being taken down because some people don't like it, including you, which goes against HN guidelines.
Hacker News biases heavily against Trump and Musk, if anything stuff like this is adored here.
No, the post got flagged and the study ridiculed because it's simply not good material for debate.
You're changing your arguments whenever I address the previous one. Have you noticed?
> Hacker News biases heavily against Trump and Musk
I imagine you're writing this only because you are biased for Musk (and/or Trump). Many people in Hacker News seek truth. Clearly, you're not the one, and you're deciding for this community that it shouldn't seek it either.
>You're changing your arguments whenever I address the previous one. Have you noticed?
Nope, I already pointed out the study is speculative bullshit: The axiom is "potential algorithmic bias around July 13, 2024".
"Potential" bias (read: changes) around one of the most significant days in the entire election cycle is why X suddenly shifted further Right from July 13 onwards?
So much happened on July 13 that changes to the algorithm (if any) at most played a minor factor, let alone "potential" changes that the study doesn't even have the gall to affirmatively declare.
Sincerely: Stop wasting my (our?) time with this bullshit.
>I imagine you're writing this only because you are biased for Musk (and/or Trump).
You are correct that I like Trump and Musk, I voted for Trump and I continue to be amazed at the crazy stuff Musk pulls off. A look through my comment history would indicate that, too.
>Many people in Hacker News seek truth.
If by "truth" you mean primarily Left/Progressive "truth". I will break the guidelines by saying this, but the audience here really isn't that different from Reddit with regards to politics.
This is to be expected, of course: Hacker News is a forum for tech enthusiasts/professionals and small cap investors operated by venture capitalists, and most if not all of those demographics are heavily Left/Progressive.
One thing I will point out that I think is fairly unique to Hacker News is the sheer level of resentment towards Musk. You can't even have a conversation about the guy unless it's to shit on him. If most other communities suffer Trump Derangement Syndrome, Hacker News suffers Musk Derangement Syndrome.
I admire that you've confirmed your biases.
By truth I mean scientific discourse centered on evidence and freedom of speech. Unfortunately, Musk's and Trump's "freedom of speech" includes freedom to disinform the public to achieve their goals, so it isn't equivalent to everyone's freedom of speech. By taking down this post, you're acting against everyone's freedom of speech.
The study is honest, because it doesn't make any conclusions, just provides new data analysis, potential interpretations, and a call for further research. You haven't responded to my arguments, e.g., about how the study takes into account external events.
It sounds like you're misperceiving the post because of your personal biases, but at least you're upfront about it, which I appreciate. You're taking down the post, because you don't like it. That's against Hacker News guidelines. This kind of behavior needs more attention, just like the behaviors of Musk and Trump. We need more transparency, not less, so I appreciate that you're upfront.
BTW, I haven't voted for Harris nor Trump. My perfect (non-existing) candidate would take the best from each of them. While I appreciate various things about Trump and Musk, I'm critical about the fact that they both disinform the public to achieve their personal goals. That's unethical.
>By taking down this post, you're acting against everyone's freedom of speech.
As far as I understand flagging on Hacker News, everyone's free to flag anything and becomes [flagged] upon receiving enough such flags. So it's not just me or the guy who posted about flagging, enough people said this thread wasn't worth Hacker News's time.
>The study is honest, because it doesn't make any conclusions,
That makes it worse because that means I spend my time to achieve nothing. I don't need a rocket scientist to tell me something obvious happened for speculatory allegations.
>just provides new data analysis,
Musk has been quite clear about siding with the Right/Conservatives/Liberals/Republicans when it comes to his politics, and X being his personal property is very well known to lean that way at this point.
>potential interpretations, a call for further research.
A study of knowns like this one should put forth a declaration in the absolute.
>how the study takes into account external events.
The study alleges a potential change in algorithm based off certain posts' performance metrics and Musk's endorsement of Trump. Essentially arguing correlation is causation and without even declaring there was causation to merit the study. That is bullshit, I ask the study and those who disseminate it to please stop wasting my time.
>You're taking down the post, because you don't like it.
I flagged this thread because it is not conducive to thoughtful and insightful debate. The study refuses to declare any conclusions or accusations, just potentials and maybes, and the logic is flawed from the outset because it strives to draw a desired conclusion ("there was (maybe) algorithm changes") first and then the steps to get there at any cost.
Even aside that, this thread is also about politics surrounding celebrities which are by themselves also against guidelines.
This thread and study could have just as likely been about Zuckerberg and Facebook or Spez and Reddit or dang and Hacker News, I would have still flagged it all the same.
>I'm critical about the fact that they both disinform the public to achieve their personal goals. That's unethical.
It's also unethical to publish a study that throws out allegations without merit for the express purpose of driving narratives, like this one.
Can the study cite how, where, and when X changed its algorithm? If not, using the term "potentially" to avoid outright libel is cowardly; anyone lodging accusations should have the decency to own it. Does the logic in the study make sense? Of course not, correlation is not causation is one of the first rules of logic most of us should learn and respect.
That's why I flagged this thread and study. It's worthless, it's flawed, and it's unethical and cowardly.
You continue falling to the straw man fallacy by bringing over and over the correlation is causation criticism, but the study doesn't make causal claims...
I understand that an analysis like that is not conducive to thoughtful and insightful debate for closed-minded people who do not care, or even oppose, to learn things that don't align with their interests.
You continue to ignore my responses to your points, e.g., about the fact that the study takes care of external events by applying difference-in-differences methodology, and continue writing accusations without being specific.
If anything, disinformation and censorship of research are worthless, flawed, unethical, and cowardly.
This is interesting, but let's not have a takeaway that the problem is isolated to bad actors purchasing social networking sites. Even on supposedly neutral platforms like YouTube or Facebook you'll see bad actors successfully manipulate the algorithmic process to help shape opinion.
I find it interesting to reflect on the fact that -- despite being raised on a diet of science fiction scenarios about dystopias -- my generation built these machines. In hindsight, it's blindingly obvious that algorithms would be manipulated by the wealthy and powerful to service their agendas. Of _course_ they would, every incentive was plainly there from the start. But we figured it wouldn't be a problem because it was us who built them, and we weren't the baddies.
I say this not to point the finger at other people. I worked on a major social media system -- and I was one of those who defended the system as it was being created. And we all should have known better.