Hate speech on social media is unfortunately not going away. As such, a group of researchers decided to do something with it, and using the data provided by over 150,000 geotagged tweets, created a map of where the hate is coming from.
The data was collected by the DOLLY (Digital OnLine Life and You) Project; the group, Floating Sheep, describes DOLLY as “a repository of billions of geolocated tweets that allows for real-time research and analysis.”
The team had previously documented that, unsurprisingly, hate speech online is most likely to originate in places where such sentiments are held in the offline world. To learn more about the geographic origins of tweets containing homophobic, racist, or ableist slurs, Floating Sheep decided to analyze all geotagged tweets originating in the United States, captured by DOLLY, between June 2012 and April 2013.
The group’s first hate-tracking project, shortly after the 2012 presidential election, mapped out the origin of tweets referring to President Obama with racial slurs. Though they received some criticism regarding how they analyzed the data, the resulting map’s findings were consistent with areas that are known to have a history of racism.
In this project, they addressed some of the previous criticisms by having students rather than an algorithm analyze each tweet. Where a program would treat each usage of, for example, “queer” the same, the people on the project were able to determine when it was being used positively, negatively, or neutrally; thus, the researchers could say with confidence that only tweets with clearly hateful intent were included.
The final tally was over 150,000 geotagged tweets containing intentional slurs. They were grouped by county, and then compared to the total number of tweets from each area, revealing locations with a disproportionately high usage of individual hate words.
To give one example, while Orange County, California has the highest total number for many of the slurs, in comparison to the overall number of tweets originating from the area, the percentage of tweets determined to be hateful was fairly small.
Though tweets using slurs could be found in many of the predictable areas, the researchers noted that, “tweets referencing ‘nigger’ are not concentrated in any single place or region in the United States; instead, quite depressingly, there are a number of pockets of concentration that demonstrate heavy usage of the word.”
Floating Sheep plans to test against area demographics at a later time to see what effect it has on results.