Digital Pink Slime: Measuring, Finding, and Countering Online Threats to Local News

In the past twenty years, journalism has had to evolve to keep up with the digital era that publishes stories not on print mediums but on online websites that are shared via social media in order to be consumed by the public. Unfortunately the smallest of journalistic outlets- those serving local communities- have been hit the hardest. Many local newspapers have closed due to budget cuts, but the trust in local news reporting remains high. In their place, some unsavory actors have decided to exploit this trust to share national messaging under the guise of local news. They have created hundreds of websites designed to appear as part of small American communities particularly communities in swing states and those of national electoral importance. With little to no actual reporters, these sites are largely filled with automated reporting on community budgets, weather, and sports. The primary pull for these websites are their partisan political articles which are shared on Facebook, Twitter, and Reddit.

This thesis is a comprehensive study into these sites that are masquerading as local news while pushing a national agenda and spending significant money to do so. Each element of this research aims to answer the questions: how is the behavior of those creating and sharing pink slime sites different from that of other news sites (be it local, real, or low credibility news)? Furthermore, how can I leverage their defining characteristics to train others to find and be wary of these sites? I start by concretely defining this phenomenon of ‘pink slime’, how it gained footing in the online news ecosystem, and what gaps in current literature inform the research I conducted. I then utilize computational social science and social network analysis methods to quantify the characteristics of these sites (in comparison to real news, local news, and low credi bility news sites) in the first large scale empirical assessment of pink slime. From an information operations perspective, I categorize the network and narrative BEND maneuvers utilized by pink slime across multiple social media platforms and compare it to three other news types across social media platforms. Applying natural language processing, machine learning, and network analysis, I propose a network feature that can find new sources of these sites and prove its effec tiveness. In a study on human subjects, I learn how a reader’s trust in pink slime and local news differs and how training impacts their ability to recognize pink slime. Finally, I summarize the f indings and relevant literature to make policy recommendations to counter this threat to local communities.