NEWS LOVES COMPANY: INCIDENTAL NEWS CONSUMPTION, CONTEXT, AND ASYMMETRIC POLARIZATION

Abstract

Research on fact-checking often asks whether corrections change what people believe. This dissertation raises a different question: how do the meaning and shareability of a correction change when it is encountered incidentally, alongside partisan content, in a feed that the reader is browsing for other reasons? Standard models of media effects assume that a message carries a fixed meaning and is processed deliberately, one unit at a time. However, these assumptions do not apply to algorithmically curated feeds, where posts arrive in a diffuse manner and attention is partial. In this dissertation, I develop the concept of “meaning leakage”: the process by which evaluative cues from an adjacent post migrate to a focal post, altering what its sharing would signal without changing its content. I argue that the mechanism operates in the evaluative register, not the epistemic one. The dependent variable throughout the study is the willingness to share, treated as evaluative acceptance rather than belief revision. I test this using survey experiments conducted in Brazil during the 2022 presidential election (approximately 2,400 respondents) and in Argentina during the 2021 legislative elections, and I interpret the results through twenty-two in-depth interviews with Brazilian voters. The mechanism operates in both countries, but applies to different objects. In Brazil, a fact-check about vaccines alongside a partisan post suppresses sharing, and the effect varies by party: Lula’s voters experience evaluative ambiguity, reducing the sharing of content from their own group without rejecting the correction, while Bolsonaro’s voters disengage themselves from the correction itself and remove it from circulation. In Argentina, where no fact-checking institution was perceived as a partisan adversary, the rival politician shapes the response, political proximity amplifies sharing, and personal restraint coexists with high expected circulation. The interviews establish that voters reason within the evaluative framework posited by the model and reframe the response of Bolsonaro’s voters, shifting from a defensive reinterpretation to a more heterogeneous evaluative disengagement. My argument has implications for the design of platforms and fact-checking: when placed alongside partisan content, a correction inherits its politics, and a tool intended to reduce polarization may actually deepen it.