The Generalizability of the Agreement Effect: A Topic Sampling Approach to Understanding Interpersonal Impressions

Social media enables people to form rapid impressions of others based on minimal information, often amplifying division and polarization. This research examines how the agreement effect—the tendency to like those with shared opinions and dislike those with opposing views—operates in this digital context. While prior work has extensively documented this phenomenon, research has largely relied on contrived person perception scenarios, limiting generalizability. Additionally, existing studies primarily assess similarity based on multiple opinions, whereas social media interactions often reveal only a single expressed belief. To address these gaps, we employ a topic sampling design, systematically testing the agreement effect across a large and diverse set of real-world topics. Using the Database of Attitude Objects (DAO), we randomly sample attitude objects with normed properties (e.g., moral relevance, emotionality) to examine how content-specific factors shape the effect. Our findings reveal not only robust evidence for the agreement effect in minimal-exposure settings but also substantial heterogeneity in its strength depending on the topic’s characteristics. This research advances both social psychological theory and methodological rigor, offering a scalable blueprint for improving generalizability in behavioral science. By identifying conditions that exacerbate or mitigate social division, these insights have important implications for digital discourse and online polarization.