Non-credibility scores: Relative ranking of news sites shared on social media to identify new pink slime sites

Pink slime news sites are politically polarized Web sites controlled by partisan national organizations that masquerade as local news. Instead of authentic community reporting, these sites rely on automated algorithms and APIs to fill in news articles between their politically charged messaging aimed at influencing votes. Over 1,000 of these sites have been identified, and the creator of the largest pink slime organization, Metric Media, has a goal of adding 15,000 more. Current methods of discovering new pink slime sites remain challenging and involves a tedious IP address lookup process to find new sites within an already existing network. This research proposes a semi-supervised learning methodology for detecting emerging pink slime sites within Facebook groups. It develops a non-credibility score as a metric to represent the trustworthiness of a news domain. With assigned scores to news domain, this research then analyses the network analysis of the Facebook pages that share content linking to pink slime sites. The non-credibility score allows researchers to efficiently survey the social media landscape to find new sources of pink slime as they emerge within the U.S. news landscape. This paper demonstrates the importance of a non-credibility score as a measure to determine the credibility of a news site through machine learning validation and then applies the methods to a recent dataset for the discovery of new pink slime sites.