> right-wing content was shown most prominently, regardless of users' political leaning.
> Left-wing users saw almost the same amount of left-wing and right-wing content, even though they only followed left-wing accounts. But only 14% of the political content sent to our right-leaning users was left-wing.
> The neutral users saw twice as much right-wing content as left-wing content. And barely any of the political content shown to our X users was non-partisan.
> Over half of the political content shown to our new X users came from accounts categorised as extreme. Of the content from extreme authors, 72% of posts came from right-leaning accounts.
So journalists ran a study with extremely poor sampling, base rate, and classification methodology designed by left-wing researchers and activists groups (including groups explicitly opposed to X), found the conclusions they wanted to find, and spent the second half of their article decrying the lack of heavy-handed moderation that pre-Musk Twitter used to control speech.
All we can actually derive from that article is that X’s new user recommendation system exposed a small number of near-default, low-feedback, synthetic accounts to more content that left-wing “misinformation” researchers classified as “right-wing”.
I'm not sure where you got any of that from. Where are you reading anything about the researchers working for Sky News, a Comcast subsidiary (!) being "left wing"? What are you basing your criticism of sampling and classification on?
I provided a study that literally won the Royal Statistical Societies' "Statistical Excellence in Journalism" award [0]. You provided unfounded dismissal and conspiracy. The burden of proof is on your side.
algo is what can make or break a propaganda machine.
Here's a pretty in-depth investigation into the workings of X as far-right propaganda machine: https://news.sky.com/story/the-x-effect-how-elon-musk-is-boo...
> right-wing content was shown most prominently, regardless of users' political leaning.
> Left-wing users saw almost the same amount of left-wing and right-wing content, even though they only followed left-wing accounts. But only 14% of the political content sent to our right-leaning users was left-wing.
> The neutral users saw twice as much right-wing content as left-wing content. And barely any of the political content shown to our X users was non-partisan.
> Over half of the political content shown to our new X users came from accounts categorised as extreme. Of the content from extreme authors, 72% of posts came from right-leaning accounts.
So journalists ran a study with extremely poor sampling, base rate, and classification methodology designed by left-wing researchers and activists groups (including groups explicitly opposed to X), found the conclusions they wanted to find, and spent the second half of their article decrying the lack of heavy-handed moderation that pre-Musk Twitter used to control speech.
All we can actually derive from that article is that X’s new user recommendation system exposed a small number of near-default, low-feedback, synthetic accounts to more content that left-wing “misinformation” researchers classified as “right-wing”.
Musk also uses plenty of heavy handed moderation; it's just almost all directed at the left. For instance: https://crimethinc.com/2022/11/25/elon-musk-bans-crimethinc-...
4 replies →
I'm not sure where you got any of that from. Where are you reading anything about the researchers working for Sky News, a Comcast subsidiary (!) being "left wing"? What are you basing your criticism of sampling and classification on?
I provided a study that literally won the Royal Statistical Societies' "Statistical Excellence in Journalism" award [0]. You provided unfounded dismissal and conspiracy. The burden of proof is on your side.
[0] https://rss.org.uk/news-publication/news-publications/2026/g...
1 reply →