Fig 1.
Y-axis values are estimated values of the total number of posts containing hate speech in a given week.
Fig 2.
Mean likes and reposts of (a) hate posts and (b) baseline posts before and after Musk’s takeover.
Black vertical lines represent standard errors.
Fig 3.
Increases in hate speech for different dimensions of hate.
Black lines represent standard errors.
Fig 4.
Coordinated networks’ largest components using three different metrics.
(a) hashtag sequence [32], (b) co-repost [32, 33], and (c) activity [32, 33]. In each panel colors are represent nodes (accounts) associated with each metric of coordination. Edges connect nodes if the two accounts are sufficiently similar. The most common non-numerical hashtags are shown to indicate major topics of discussion within each network.
Fig 5.
Changes in coordinated and non-coordinated account activity and interactions after Musk’s purchase.
(a) Percent change in posts per day, (b) change in mean reposts per post, and (c) change in mean likes per post before versus after Musk’s purchase. Black bars represent standard errors. We use 3 different coordination metrics: hashtag sequence [32], co-repost [32, 33], and activity [32, 33].
Fig 6.
Bot scores for random accounts before and after Musk’s purchase (bot scores vary between 0, organic users, and 5, clear bot activity).
Fig 7.
Cashtags become substantially more popular after Musk purchased X.
(a) Most frequent cashtags (inset is an example post). (b) Decrease in cashtag frequency before Musk purchased X, and substantial increase after Musk’s purchase.