Opinion Dynamics in Online Social Media

Social media have become an important source of information for many of the billions of people who use internet on a regular basis. Pundits and scholars have warned for various adverse effects that its use could have on opinion formation. Personalization algorithms would create filter bubbles—individual information environments in which users become more and more convinced of their own beliefs—and social bots freely spread misinformation to a gullible audience.

This dissertation aims to critically evaluate the validity of folk-theories on opinion dynamics in online social media and provide a lens for understanding the ways in which social media platforms shape the process of opinion formation. Using theoretical agent-based computational models and online experiments I investigate three different paths to polarization: the role of personalization algorithms, the strategic use of social bots and the mass dispersion properties of content on social media. First, I find a mismatch between the role of personalization algorithms and state-of-the-art social influence models. Using the information from an experimental study with Facebook users, I predict that personalization of online social media could prevent polarization, particularly when distant groups use different moral foundations of argumentation. Second, the fact that bots stand largely disconnected from human users on social media might make them more, not less effective in spreading their content. Third, the rapid sharing of information to all your contacts at once—typical for information sharing on platforms like Facebook and Twitter—may contribute to isolation of individuals who hold slightly different beliefs.