Skip to main content
. 2020 Oct 20;3(2):445–468. doi: 10.1007/s42001-020-00087-4

Fig. 2.

Fig. 2

Normalized mean centrality of users in the datasets. Users are divided into low, medium, and high hate scores, respectively defined as the bottom quartile, the middle 50%, and the top quartile of hate scores. Bot predictions are based on an 80% probability threshold