Figure 6

Impact of hate and counter focal tweets on subsequent discourse over time. Each colored square represents the average difference between tweets following and preceding a focal tweet posted in a reply tree in a particular week (Eq. (2)). Results are computed for 82,132 reply trees with at least 3 tweets, over time (horizontal axis), and are binned by hate-score of the focal tweet (vertical axis). Red (blue) squares signify that the focal tweet was followed by a change in the subsequent conversation towards more hateful (counter) speech, with color saturation corresponding to the size of the difference. Event labels are the same as in Fig. 3