Trust, Truth, and Selective Exposure to Partisan News
Abstract
We will present preliminary pilot results from a project designed to measure the psychological mechanisms behind ideological segregation in digital news consumption. Many commentators (for example, Cass Sunstein in his book Echo Chambers) have raised concern that Americans are increasingly exposed to news and information that confirms instead of challenges their pre-existing ideologies. Sunstein and many others fear that this will make it increasingly difficult for democracies to function, and the events of the last two years in the U.S. and abroad only underscore this concern. A large body of literature empirically documents the extent of this ideological segregation and points to demand, not supply, as the main driver.This raises a key question: why do we tend to demand ideologically aligned news? Theory and laboratory evidence have highlighted a range of mechanisms that could be at play, but field evidence on their relative importance is limited. The existing theories can be broadly grouped into two categories: (1) those in which partisans believe ideologically aligned sources provide more accurate information; (2) those in which partisans enjoy confirmation of their beliefs, or partisan “cheerleading,” and so choose ideologically aligned sources despite believing they may provide less accurate information.
To distinguish these mechanisms, we build on the following insight: when the stakes are high, we turn to outlets we trust to provide accurate information. Thus, we can identify what outlets people trust by asking people to make predictions about verifiable future events, and randomly varying the incentives for correct predictions. Specifically, we have designed a field experiment where subjects are asked to make predictions on events during the upcoming month (e.g. the occurrence of deadly terrorist attacks). Before making such predictions, subjects are asked to choose news outlets and articles from a news feed that is similar to Facebook or Google News. We randomly provide monetary incentives to a subset of the subjects to reward their correct predictions. In addition, we randomly inform a subset of the subjects regarding the accuracy of the news outlets that they can choose from. The degrees to which subjects respond to incentives and information allow us to disentangle utility driven by a desire for accuracy from psychological utility that is unrelated to accuracy.