After the 2015 Brexit Referendum in the United Kingdom and the 2016 presidential election in the United States, so-called “fake news” has been a central component of political discourse. Fake news stories have proliferated online through social media platforms like Facebook and Twitter and the abundance of fake news has led to numerous congressional hearings on misinformation trying to determine the power of fake news. If we assume that people believe fake news stories and assume that people will propagate fake news stories on social media in the same way they spread real news stories, then it follows that misinformation is bad for democracy because the fake news stories create more uncertainty for voters.
In her November 5th talk at Northeastern University’s Network Science Institute, “How Bad is Fake News? Motivations for Sharing Misinformation Online,” visiting speaker Professor Miriam Metzger discussed her recent research into why people share fake news on social media, examining whether or not people believe the fake stories in the first place. If someone does not believe a fake news story, then the motivation for sharing will likely be different than the motivations among those who believe the story is true. For example, motivations may include sarcasm, incitement, sociability, or financial gain. In her talk, Metzger outlined a program of studies that work to discover the variety of motivations for sharing fake news and the believability of the news stories.
One study looks at the news content of tweets surrounding misinformation spread in Twitter, using the replies and retweets of all tweets that were fact checked by Politifact between January 1st and May 3rd, 2019 determined to be “false” or “pants-on-fire” on the Politifact scale. The replies and retweets were then coded and evaluated for belief or disbelief. However, the methodology does not capture second-order replies and retweets by subsequent Twitter followers. Metzger and her collaborators find that highly emotional and partisan issues—such as tweets about House Speaker Nancy Pelosi and about immigrants—tend to be more believed. The study also finds that there is more disbelief expressed by the followers of President Trump and more belief by the followers of partisan political pundits. Metzger and her collaborators hypothesize that the broader social media base—that is followers–of the President includes many liberals in addition to his conservative base. Pundits, however, have a larger ratio of like-minded followers, possibly indicating confirmation bias whereby people are more likely to believe information if it confirms their political opinions. Therefore, a like-minded source combined with a highly partisan or emotional issue is the most dangerous form of Twitter misinformation.
Another study Metzger discussed is an attempt to develop a computational model for detecting disbelief with the goal of creating a dataset that can automatically classify tweets, replies, and retweets as reflecting belief or disbelief in misinformation spread on Twitter. Using two different models of natural-language processing—a bag of words model combined with a support vector machine model, and a long short-term memory model—Metzger and her team have achieved between 69 and 73% classification accuracy so far. The next steps are to refine their model to enhance accuracy so it can be used in studies that aim to understand the degree to which people believe, or disbelieve, political misinformation.
While Professor Metzger’s goal of determining the motivations for sharing misinformation is still ongoing, the studies she discussed on November 5th have resulted in important findings and have created new, testable hypotheses for understanding the impact of fake news. While some people share fake stories because they believe them, this is not always the case. Furthermore, in many cases there is a great deal of disbelief of fake stories and some sources of news stories are believed more than others. The motives behind misinformation sharing are nuanced and complicated, but more research into the phenomenon of fake news can help us better understand the true behaviors at work in our misinformation-rich political environment.