The contingent reputational benefits of selective exposure to partisan information.

Abstract

Individuals often preferentially avoid information that contradicts and seek information that aligns with their prior beliefs-a tendency referred to as "selective exposure." Traditionally, prior research has focused on intrapersonal drivers of selective exposure, including avoidance of cognitive dissonance. We take a complementary approach by investigating the conditions under which interpersonal concerns drive selective exposure. Drawing on a large literature on impression management, we test a social signaling model of selective exposure, which predicts that (a) individuals shift their information selection decisions to signal to observers and (b) observers reward such shifts. We test this model in the domain of partisan politics in the United States across five financially incentivized, preregistered experiments (N = 3,598). Our results extend prior theory by identifying three key contingencies: the type of task on which observers expect to collaborate with actors, alignment of group membership between observers and actors, and the magnitude of demonstrated selective exposure. Overall, we find that tailoring one's information selection decisions can indeed have strategic value-but only under certain theoretically predictable conditions. Our work also identifies an actor-observer misalignment: While observers are sensitive to the type of future interaction with an actor, the actors themselves do not intuit this sensitivity. In the era of social media, when information selection decisions are more public than ever and the spread of misinformation is pervasive, understanding the ways in which reputational considerations shape decision making not only illuminates why selective exposure persists, but also suggests novel mitigation strategies. Public Significance StatementIn the domain of partisan politics, we provide experimental evidence for the interpersonal drivers of selective exposure to information. We find that people shift their information selections to pander to observers and that observers subsequently reward people who select information aligned with the observers' ingroup.