Engaging politically diverse audiences on social media
News Bridge studied how the language that a media outlet – PBS’ Frontline – uses to promote its content influences the political diversity of its audience. In this project we tracked user engagement with tweets posted by Frontline over three years and built models that, given the tweet text, predict the political diversity of the audience. We then integrated the models into a web app that helped Frontline craft tweets engaging to a politically diverse audience, guided by the model predictions. While studies of political polarization on social media typically investigate the behaviors of individual users, New Bridge focused on the media outlets’ impact on audience fragmentation and developed tools that can help them reduce it. We believe this approach can be further developed and generalized into tools to help communicators (e.g., in public health) engage audiences across political, social, and/or cultural boundaries.
Motivating Context
The U.S. news media is more politically fragmented than ever, with Americans of different political identities inhabiting divergent media worlds, and with ever more separation between the sources of information that they engage with and trust. Media outlets have an opportunity to counteract this polarization by actively promoting their content in a way that brings in a more politically diverse audience. In this context, News Bridge was motivated by two research questions:
1. Can we build models that predict the political diversity of an audience that will engage with a tweet given the tweet text?
2. Can we use such predictive models to help media outlets select tweets that are more likely to engage a more politically diverse audience?
Approach
With News Bridge, we explored how political polarization is reflected in the social media posts used by media outlets to promote their content online. In particular, we tracked the Twitter posts of several media outlets over the course of more than three years (566K tweets), and the engagement with these tweets from other users (104M retweets), modeling the relationship between the tweet text and the political diversity of the audience. We then built a tool that integrated our model and could help a media outlet or journalist craft tweets that are engaging to a politically diverse audience, guided by the model predictions. To test the real-world impact of the tool, we partnered with the PBS documentary series Frontline and ran a series of advertising experiments on Twitter. We found that in seven out of the ten experiments, the tweets selected by our model were indeed engaging to a more politically diverse audience, illustrating the effectiveness of our approach.