Conversation

Sorry that the previous Y axis scale shifted between the comparisons. Here they are side by side with the same scale.

This is showing how the toxicity level doubles on #gab when comparing the average of all posts with the average of posts only containing the words "jew, jews and jewish." #dataviz

2:39 PM · Dec 6, 2018Twitter Web Client

Replying to

Note that, at least in previous version of API, simple identity markers scored higher on toxicity, making for a biased distribution of false positives. Thus, it's possible that simple statement of Jewish identity or pro-Jewish posts will be misclassified engadget.com/2017/09/01/goo

Replying to

and

That said, I don't have high hopes for Gab w/r/t antisemitism.

Replying to

Keep working on this, please! Can you show graphs for some of Gab's other favorite targets: Blacks/African-Americans; Muslims; Dems/Libs; ordinary conservatives ("cuckservatives"). How prevalent are posts targeting these groups and what is the relative score in each case.