X's algorithm was changed in mid-July 2024 to boost right-leaning accounts | Hacker News Reader

X's algorithm was changed in mid-July 2024 to boost right-leaning accounts

mastodon.social

89 pointshn198650 comments

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.

Yeah like an assassination attempt.

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

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!!!

This could be the result of the news feed change at X and Musk's endorsement of Trump...

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.

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

> 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.

Premium accounts get a boost.

You'd struggle to convince me the number of Premium registrations wasn't heavily biased to the Right.

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"

So because of the shooting Musk endorses the orange idiot plus has the algo modified. Sounds plausible.

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

I have no idea why this post is being downvoted and nobody else has bothered to reply and actually provide substantial objection.

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.

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

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

wow, "view count findings suggest possible recommendation algorithm bias", much science

So, election interference?

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.

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?

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 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.

The most important guideline of all; thou shalt not make weird billionaires look bad.

Thanks for pointing to the guidelines, I didn't realize I was the one breaking them by commenting that I flagged this. But there is "Please submit the original source. If a post reports on something found on another site, submit the latter. "

I have a feeling that the research on its own would not have generated the discussion that the toot did, because the toot really just prompted people to give their own impression without discussing the actual research. Sometimes I flag stuff just because it's flame bait, maybe the submission didn't break guidelines but the comments will be snarky and shallow and not conducive to building the community culture. And I rely on the democratic aspect of HN, if my flag is frivolous so be it, no harm done.

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.

Reply on news.ycombinator.com