Running While Female: Using AI to Track how Twitter Commentary Disadvantages Women in the 2020 U.S. Primaries

Source: Sarah Oates, Olya Gurevich, Christopher Walker, Lucina Di Meco, Philip Merrill College of Journalism – University of Maryland, The Wilson Center, and Marvelous AI, August 28, 2019

From the abstract:
While there is conclusive research that female political candidates are treated unfairly by traditional media outlets, the volume and pace of information flow online make it difficult to track the differentiated treatment for female candidates on social media in real time. This paper leverages human coding and natural language processing to cluster tweets into narratives concerned with policy, ideology, character, identity, and electability, focusing on the Democratic candidates in the 2020 U.S. Presidential primary election. We find that female candidates are frequently marginalized and attacked on character and identity issues that are not raised for their male counterparts, echoing the problems found in the traditional media in the framing of female candidates. Our research found a Catch-22 for female candidates, in that they either failed to garner serious attention at all or, if they became a subject of Twitter commentary, were attacked on issues of character and identity that were not raised for their male counterparts. At the same time, women running for president received significantly more negative tweets from right-leaning and non-credible sources than did male candidates. Following the first Democratic debates, the individual differences between male and female candidates became even more pronounced, although at least one female candidate (Elizabeth Warren) seemed to rise above the character attacks by the end of the first debates. We propose that by using artificial intelligence informed by traditional political communication theory, we can much more readily identify and challenge both sexist comments and coverage at scale. We use the concept of narratives by searching for political communication narratives about female candidates that are visible, enduring, resonant, and relevant to particular campaign messages. A real-time measurement system, developed by MarvelousAI, creates a way to allow candidates to identify and push back against sexist framing on social media and take control of their own narratives much more readily.