I use computer science to test social science. Social science theory predicts behaviors. Social media provide behaviors of millions of people, businesses and organizations. Computer science allows us to scale content analysis to big data proportions.

I am one of a few social scientists in communication that can not only understand advanced computer science methods, but actually apply them. There is no reason we can’t build predictive models to solve real-world problems. As communication scholars, we can take millions of messages and process them to detect when events occur, or when people are acting a certain way. We then can make predictions on their behaviors based on this knowledge.

Recently, I published a study with Ted Tomeny at the University of Alabama where we looked at how vaccine talk on Twitter (specifically the myth that vaccines cause autism) changed over time. Here’s an animated map that shows how that talk changed across time. The darker the shade, the stronger anti-vax sentiment expressed in that area.

In the article (which you can find here), we not only detect and map vaccine sentiment using machine learning, we also use census variables to predict the occurrence of anti-vax behaviors. Here we use variables that fellow social scientists have shown are linked to anti-vax sentiment, such as higher socioeconomic status, and assess their ability to explain what we observe on the map.

To investigate how pro (or anti) vaccine your geographic area is, check out the map with overall area averages (across 5 years) below.

Having issues with the maps? Try the fullscreen version of the animated version here, or the overall map here.

Last year, I gave a presentation on the varying ways in which I have looked at millions of messages on Twitter. The majority of these studies can be found in the Publications section of this website.

Download the PDF file .