Characterizing Information diffusion. Social influence, propagation speed, polarization

Abstract

Social-media platforms have created new ways for citizens to easily access information and participate in political discourse. However, social media with its algorithmically curated and virally propagating content contributed further to the polarization of opinions by reinforcing users existing ideological viewpoints: users favor narratives that confirm their claims. Hence, identifying the ideological leanings of social media users and how those affect their information diet is a first step, of crucial importance, towards devising techniques for mitigating confirmation bias. Content is delivered to users of a social-media platform mainly through two mech- anisms: (i) via the content- recommendation engine of the platform, or (ii) via social connections of users, e.g., friends or followees, who share, comment, or like relevant content items. In this talk we focus in the latter mechanism of content delivery, i.e., content seen in ones feed due to social connections. In particular, we are interested in modeling the propagation of content in an online social network, through the lens of ideological leaning. Identifying the ideological leaning of social-media users is a crucial step for better understanding echo chambers, conceiving tools to mitigate their effects, as well as designing public information campaigns. We will review the existing approaches to the identification of ideological leaning of users and content, as well as the study of information propagation from the ideological standpoint. We also illustrate a novel steps in this direction that propose models to jointly learns users ideological leanings and the tendency of politically salient content to propagate from one user to another from homophily-driven user interactions.