As part of Stanford's Democracy Day 2024 series of events, Stanford Data Science hosted a discussion about the intersection of data science and politics.
Scholars, Ruth Appel, Ross Dahlke, and Ryan Moore previewed their recent papers and explored how data science can be used to understand the role of deceptive influence campaigns on social media and misinformation in democracy. Their papers explored the challenges faced by social media companies in addressing deceptive influence campaigns, how misinformation spreads on the internet, and distinguishing features between misinformation and other types of content, such as high-quality news and “pink slime” journalism.
More specifically, Ruth motivated her work by sharing that there is little large-scale research that analyzes the reach and impact of deceptive influence campaigns. This gap in understanding leaves us debating their effects without clear answers.
Driven by a desire to understand how these campaigns shape democracy, Ruth’s paper dove into the characteristics, reach, and impact of deceptive influence campaigns that reached U.S.-based Facebook and Instagram users between June 26, 2020, and February 15, 2021. She analyzed how these campaigns were crafted, what kind of content users were exposed to, and how people interacted with them. By examining these factors, Ruth shed light on what influence campaigns looked like in practice, who they reached, and how users engaged with them. Findings are crucial for understanding the potential of these campaigns to impact democratic dialogue online—and ultimately, to explore what might be done to counter their effects.
Ryan shared the findings of a study about pink slime journalism. Pink slime journalism refers to organizations that pose as genuine local news outlets but often publish content driven by special interests. While past research has shed light on how these outlets operate—like identifying them and documenting their tactics—no one has dug into how people consume this "pink slime" content.
In his paper, Ryan analyzed the browsing habits of a representative sample of American adults during the 2020 election season. The sample includes data from 1,238 people, who together visited a whopping 21 million websites. His team matched this data with a list of 1,627 pink slime websites, containing over 132,000 distinct articles, to answer the key questions: who’s consuming pink slime, what are they reading, when and where are they seeing it, and what might be drawing them in.
Findings were surprising: around 3.7% of American adults—about 9.5 million people—visited pink slime sites during this period. However, the way people interacted with these sites was very different from how they consumed either traditional local news or even misinformation sites. Interestingly, there was a big mismatch between the topics these outlets produced and promoted, and the topics that visitors read. And although Google and Facebook were major drivers of pink slime traffic, people in “news deserts” (areas with limited local news sources) weren’t necessarily more likely to see this content.
Then Ross concluded the discussion by sharing his paper on contextualizing misinformation exposure, delving into what people read by analyzing the topics and linguistic features of misinformation that users were exposed to during their web browsing. Using web data from a representative panel of 1,240 American adults during the 2020 U.S. Presidential Election, his team analyzed the linguistic patterns in over 107,000 unique misinformation and traditional news pages.
Ross observed that misinformation content tends to be easier to read, shows more negativity, and uses more moral language than traditional news. Interestingly, misinformation isn’t one-size-fits-all—its tone and language vary widely depending on the topic. His team also found big differences based on who was reading: older adults, for instance, tended to consume more COVID-19 and health-related misinformation, which leaned more negative in tone. Republicans, on the other hand, engaged with misinformation that was more negative and morally charged and focused on social and political topics rather than health.
Ross’s findings underline the importance of understanding misinformation through a user-centered lens. The findings suggest that combating misinformation effectively will require targeted approaches that account for both the topics people engage with and their unique characteristics.
Useful Links
The Consumption of Pink Slime Journalism: Who, What, When, Where, and Why? (Pre-print paper); authors: Ryan Moore, Ross Dahlke, Priyanjana Bengani, and Jeffrey Hancock
Scenes from Democracy Day 2024 (Stanford Report coverage)
Stanford Democracy Day Team—Kudos to the organizers and all the volunteers!
