Geography of Hate: mapping racist and homophobic tweets

We can use social media to predict illnesses, but can we interpret social media well enough to accurately understand the thoughts and biases of its users?

Enter the Geography of Hate project.

More than 150,000 geocoded tweets sent between June 2012 and April 2013 were analyzed by a team of students at Humboldt State University in California to produce the Geography of Hate heatmap, which uses the Google Maps API to categorize from what counties in the U.S. homophobic and racist tweets are sent, as well as tweets disparaging people with physical disabilities.

View the Geography of Hate map here.

More about the methodology behind the project is here, but the map itself is fascinating for its attempt at sentiment analysis.

• Since sentiment analysis based solely on algorithms would classify any tweet containing one of the chosen “hate words” as a “negative” tweet, the students involved in the project had to read each tweet and classify it based on a “predefined rubric,” according to the details.
• So, for instance, a tweet that is quoting a line that includes the N word from “The Adventures of Tom Sawyer” possibly would’ve been picked up as negative according to algorithmic analysis, versus a student’s reading the tweet and discerning what sort of meaning was intended.
• The students from Humboldt State University involved in the project therefore had to reach every single tweet and then interpret each tweet’s meaning based upon a rubric they were given, and then classify the tweets as “positive,” “neutral,” or “negative.” Only tweets classified as “negative” were used to produce the heatmap.

One alteration could have proved beneficial: revealing the ratio of Twitter users pegged for “negative” speech in each county to the total number of Twitter users per county, which might have provided a more precise representation of counties’ relative prejudices.

This project is a rather compelling reason for why courses in the humanities are ever more important in the Age of Algorithms, and certainly at a time when federal education spending and public attitude appears hell-bent on dissuading every future English literature major from that degree that will surely prepare them only for jobs as coffee shop baristas.

“Big data,” without proper interpretation, and with insufficient sorting, is nothing more than a deluge of information missing its proper context — intellectual laziness masquerading as meaningful thought because a computer program, and not the lessons of “The Merchant of Venice,” derived the answer.