Tweets Used to Generate ‘Geography of Hate’

By Cameron Scott 

Researchers at University of Humbolt State University used aggregated tweets to assemble a map of where hate speech is most prevalent.

The map is interactive, allowing users to sort by racist, homophobic and anti-disability remarks.

The results are rather damning of the Eastern United States. Admittedly, population is denser there than in most of the West, but that doesn’t fully account for the difference on the map between the two sides of the country.

To wit: Rural Oregon and Idaho have red counties, and California has almost no red at all (when viewed closer in, the hotspot of hate in the state is on the Nevada border).

Computers flagged tweets using a set of keywords, and the students went through all of the tweets by hand to ensure that the context was indeed bigoted. (Undergraduates are cheap labor!)

Counties at or under the national average are left unshaded. Those with slightly higher than average instances of hateful speech are coded blue, and those where bigotry abounds are red. All the data is based on tweets aggregated at the county level, and the map is normalized for the number of tweets in each county.

The study, organized by Monica Stephens, obtained Twitter data from the University of Kentucky’s DOLLY project and used the Google Maps API.