Hostility towards outsiders motivates engagement on social media

An analysis of nearly 3 million posts from news media accounts and U.S. congressional members on Facebook and Twitter found that people share or retweet posts about opposing political groups twice as often as posts about their own political group. The study was published in PNAS.

Social media platforms like Facebook, X (formerly Twitter), and Instagram allow users to share content, communicate, and interact in real time. For many, these platforms have become a primary source of news, but they also play a central role in how people build and maintain social relationships.

Unlike traditional media—where all viewers see the same content—social media sites use algorithms to personalize the content shown to each user. These algorithms determine what posts appear in someone’s feed, meaning that no two users see exactly the same content.

While the purpose of these algorithms is to keep users engaged by showing them content they are likely to enjoy, researchers have long warned that such personalization, combined with user behavior, can create “echo chambers”—informational bubbles in which users are primarily exposed to content that confirms their existing beliefs. While this may be harmless in contexts like music or movie recommendations, echo chambers around political content can exacerbate polarization and deepen societal divisions.

Study author Steve Rathje and his colleagues investigated whether out-group animosity—hostility toward political opponents—generates greater engagement on social media. They hypothesized that in a polarized society, expressing hostility toward opposing groups may be a more effective way of signaling partisan identity than praising one’s own group.

In their first analysis, the authors examined Facebook and Twitter accounts of U.S. liberal (e.g., The New York Times, MSNBC) and conservative (e.g., Fox News, Breitbart) media outlets. They found that each additional negative emotional word in a post was associated with a 5% to 8% increase in shares and retweets—except for posts from conservative media on Facebook, where such words actually decreased sharing. More strikingly, posts referring to political out-groups had a 35% to 57% higher chance of being shared for each additional out-group word. These posts also attracted more “angry” and “haha” reactions—across both liberal and conservative media.

The second analysis focused on the Facebook and Twitter accounts of U.S. congressional members from both the Democratic and Republican parties. Similar patterns emerged: negative emotional language made posts more likely to go viral. Each additional negative emotion word increased the likelihood of a post being shared by 12% to 45%. Posts referring to political out-groups were again much more likely to be shared, with a 65% to 180% increase in odds per out-group word. Posts attacking the out-group drew more “angry” reactions, while those referring to the in-group (i.e., the poster’s own political party) attracted more “love” reactions.

“While social media platforms are not fully transparent about how their algorithmic ranking system works, Facebook announced in a post titled “Bringing People Closer Together” that it was changing its algorithm ranking system to value “deeper” forms of engagement, such as reactions and comments. Ironically, posts about the political out-group were particularly effective at generating comments and reactions (particularly the “angry” reaction, the most popular reaction across our studies). In other words, these algorithmic changes made under the guise of bringing people closer together may have helped prioritize posts including out-group animosity,” the study authors concluded.

The study sheds light on the links between social media contents and people’s reactions to them. However, it’s important to note that whether a post is shared or receives reactions doesn’t depend solely on users. While users choose how to engage with content, it is the social media algorithms that decide which posts are shown to whom—and how frequently. If these algorithms were to change, user behavior and engagement metrics could shift accordingly.

The paper, “Out-group animosity drives engagement on social media,” was authored by Steve Rathje, Jay J. Van Bavel, and Sander van der Linden.