Individual Submission Summary
Share...

Direct link:

Estimating the Ideological Content of Images with an Application to Congress

Sat, August 31, 8:00 to 9:30am, Marriott, Maryland B

Abstract

In this paper, we develop a method for estimating the ideological content of images posted by members of Congress on Facebook and develop a deeper understanding of how members of Congress express political rhetoric through images. While much progress has been made toward measuring ideology from roll call data and from texts, to date, little is known about how political ideology is expressed through images. Using deep learning techniques to automatically predict Republican or Democratic party affiliation solely from the Facebook photographs of 114 U.S. Members of Congress, we demonstrate that predicted class probabilities from our model function as an accurate proxy of the political ideology of images along a left-right (liberal-conservative) dimension through validation with NOMINATE scores and demonstrate how our method can be used to learn about how images are used during the course of political campaigns in primary and general Congressional elections. In addition to this, our method, along with in-depth content analyses, reveals features of images which distinguish liberal and conservative visual content.

Authors