Information Diets are More Diverse in Attention than in Engagement

Background

The diversity of political information people consume is central to democratic decision-making. Social media has generated concerns that personalization algorithms and self-selected networks create echo chambers, limiting users to ideologically agreeable content. At the same time, other research shows that users often encounter opposing viewpoints incidentally through their social networks or platform recommendations.

A major challenge is measuring what people actually consume. Most platform data capture visible actions such as likes and shares but cannot observe whether users simply read content without publicly interacting with it. Survey data also suffer from recall errors, while eye-tracking studies are difficult to conduct at scale. The authors argue that attention and engagement represent different behaviors and should be measured separately to better understand online information diets.

Study

The researchers built a social media clone that displayed a standardized feed of politically balanced posts while allowing them to independently measure attention and engagement. Attention was recorded whenever participants clicked "Read more" to expand a post, while engagement was measured through likes and shares. Feed order and displayed popularity metrics were randomized so every participant viewed the same content under controlled conditions.

The study consisted of four online experiments: one conducted in the United Kingdom and three in the United States. Political posts were collected from UK parliamentary candidates and U.S. senators, then classified along a left-right ideological spectrum using an OpenAI language model with human verification. Participants browsed the feed for two minutes before completing a questionnaire measuring political ideology and demographics. The authors then compared how strongly political ideology predicted attention versus engagement with politically congruent content.

Results

The study found that attention was only weakly associated with political ideology. Across all four studies, participants from different ideological backgrounds were willing to click "Read more" on content from both sides of the political spectrum, indicating relatively diverse information consumption.

In contrast, engagement was highly partisan. Participants were substantially more likely to like and share posts that aligned with their own political beliefs. The ideological relationship between users and engagement was consistently much stronger than the relationship between users and attention across every study and in the pooled meta-analysis.

The authors conclude that relying solely on likes and shares overstates ideological segregation online. People consume a wider range of political information than their public interactions suggest. At the same time, they caution that their clone platform cannot perfectly reproduce real social media environments, where recommendation algorithms, social networks, and platform design all influence what users encounter. Nevertheless, the study introduces a scalable method for measuring attention separately from engagement and highlights the importance of considering both when evaluating online information diversity.