By Matt Peckham, TIME
(TIME) -- Skim the zoomed-out surface of Humboldt State University’s alarming “Hate Map” and you’ll encounter angry clouds of bright red framed by smears of gloomy blue, as if some giant freak storm were raining down hell across the the United States.
What you’re looking at is actually a map created by pairing Google‘s Maps API with a hailstorm of homophobic, racist and other prejudicial tweets. It’s part of a project overseen by Humboldt State University professor Dr. Monica Stephens, who, along with a team of undergraduate researchers, wanted to test for geographic relationships to hate speech.
Above the map, the words “homophobic,” “racist” and “disability” define alternate “hate storm” views, each describing a range of highly offensive terms. Click on the keywords or any of their subcategories and the map shifts, the splotches reorganizing to reflect occurrences of the selected term: Bright red areas describe the “most hate,” while light blue ones describe “some hate.”
Creating a map like this is essentially about data-plotting: In this case, HSU says the data was derived from “every geocoded tweet in the United States from June 2012 – April 2013″ that contained keywords related to hate speech. How’d HSU collect all of that Twitter data? Through DOLLY, a University of Kentucky project that maps social media according to geography, allowing researchers to then comb through the data for patterns or correlations. But what about tweets that used the keywords in a positive (that is, “critical of them”) sense? HSU’s researchers read through the tweets manually, categorizing each as positive, neutral or negative — the map only displays the tweets categorized as negative.