Experience Stern | Faculty & Research

Overview: In “Divergent Patterns of Engagement with Partisan and Low-Quality News Across Seven Social Media Platforms,” NYU Stern Professor Jennifer Allen and co-authors Mohsen Mosleh (Oxford Internet Institute) and David Rand (Cornell University) reveal that while social media platforms diverge sharply in their political leanings, one troubling pattern remains universal — users attract more engagement when they post lower-quality news than credible journalism.

Why study this now: As social media becomes increasingly fragmented, research has continued to center on single platforms—most often X (formerly Twitter) or Facebook. In response, the researchers set out to explore the broader ecosystem and how people behave across ideologically distinct spaces. Their study captures this shift from a few dominant networks to a constellation of smaller, more partisan communities—each with its own culture, audience, and algorithms (or, in some cases, none).

What the researchers found: Drawing on more than 10 million posts linking to news outlets across seven major platforms—X, BlueSky, Truth Social, Gab, GETTR, Mastodon, and LinkedIn—the team uncovered two consistent dynamics:

• The “echo platform” effect: Conservative news gains more traction on right-leaning sites like Truth Social and GETTR, while liberal content thrives on platforms such as BlueSky.
• Low-quality news, as measured by domain ratings, consistently outperforms reputable journalism across every platform studied—even those without algorithmic curation. This pattern suggests that the bias toward sensational content may stem less from platform design and more from human preference.

What does this change: The findings challenge the assumption that algorithms alone drive the spread of misinformation. Instead, they highlight the role of users themselves—what they find engaging and choose to amplify. The implication for policymakers and platform designers is clear — meaningful progress requires addressing the social and psychological roots of engagement, not just adjusting technical systems.

Key insight: “The platform’s algorithm isn’t necessarily worsening the problem. The patterns we see across very different systems—including some that don’t use engagement-based algorithms at all—suggest these are human tendencies, not just technical ones,” explains Allen. “If we want to elevate high-quality content, we have to design social media that actively rewards credible reporting, and help users learn to recognize and value such content.”

The research was recently published in the Proceedings of the National Academy of Sciences (PNAS).