Right or left, low-quality news links popular on social media

The spread of dubious headlines on social media isn't just a right-wing thing - it's a social media thing, according to new research from Cornell.

After studying millions of social media posts containing links to news stories on a variety of platforms, the team found that news shared on platforms with more conservative user bases is, on average, lower in quality. When it comes to likes and shares, they found that news aligning with the dominant political slant of a platform got more engagement - but that on both conservative- and liberal-leaning platforms, a user's posts with lower-quality news links got more engagement than their higher-quality news posts.

"If your post is in line with the norm on the platform, people engage with it more," said David Rand '04,professor of information science, marketing and management communication, and psychology in the Cornell SC Johnson College of Business and the Cornell Ann S. Bowers College of Computing and Information Science.

"It's like an 'echo platform' scenario," he said. "Many had argued that people on the right were better at getting engagement on social media, but we find that it totally depends on the platform. When it comes to the advantage of lower-quality news, on the other hand, it's happening on both sides."

Rand is senior author of "Divergent Patterns of Engagement With Partisan and Low-quality News Across Seven Social Media Platforms," which published Oct. 30 in Proceedings of the National Academy of Sciences. The corresponding author is Mohsen Mosleh, associate professor at the Oxford Internet Institute and a research affiliate at the Massachusetts Institute of Technology Sloan School of Management.

The other co-author is Jennifer Allen, assistant professor at the New York University Stern School of Business.

Rand, an Ithaca native who returned to his alma mater this year after seven years at MIT, said much of the belief that users engaged and shared false news more than true news stemmed from a 2018 paperin Science. But that research studied just one site - Twitter (now X).

"There are many studies, including a lot that we've done, looking at the sharing of information on Twitter," he said, "and that was primarily because Twitter had an API (application programming interface) that allowed academics to easily access tweets. So people studied Twitter and then made general claims about social media."

Rand and his colleagues wanted to analyze a greater cross-section of the social media landscape, so they looked at seven platforms - BlueSky, Mastodon, LinkedIn and Twitter/X, considered more left-leaning or neutral; and TruthSocial, Gab and GETTR, considered right-leaning. They were able to collect data on individual posting behaviors, posting of outside links and engagement from all of these platforms.

They chose one month (January 2024) and analyzed every post on all seven platforms - nearly 11 million in all - that contained links to news domains.

"What we wanted to do was say, 'Let's look beyond just one platform,'" Mosleh said. "We wanted to look at conclusions that have been drawn in the past by looking at one platform, and see to what extent they either generalize across platforms or vary in systematic ways."

To gauge the quality of the news site, the team used the reliability of the publisher as a proxy for the accuracy of the content. They used a ratings system released in 2023that scored more than 11,000 news sites based on accuracy by combining numerous different sets of expert evaluations. They also estimated political lean for each domain and validated those estimates against established benchmarks.

The key takeaway: Across all seven platforms, the average user received more engagement on their lower-quality news posts. This pattern is observed even on Mastodon, which doesn't use a ranking algorithm - meaning user preference, not just algorithms. promotes engagement with lower-quality news.

Overall, higher-quality news is substantially more prevalent and garners far more total engagement across posts, Rand said, but within a given user's posts, lower-quality news is engaged with at a higher rate than higher-quality news. On average, a user's posts with links to the lowest-quality sites earned about 7% more engagement per post than their posts with links to the highest-quality sites.

One important feature of their results, Rand said, is that they account for all possible variation due to characteristics of the poster. That way, they can see which type of news gets more engagement, regardless of whether the posting individual has, for example, millions of followers or just a few.

"So we can say our result is completely not about differences in the characteristics of the poster," Rand said, "but it's really just about the characteristics of the content. It's not the algorithm, it's not the user who's posting it: A user's posts with lower-quality links get more engagement."

Funding support for this work came from the Open Society Foundation.