Evaluating the impact of content warning labels on individuals with prior fringe beliefs

Abstract

Recent academic research suggests that exposure to misinformation is rare and heavily concentrated among a small group of people who already have extreme or misinformed views. Despite this evidence, research on content interventions, such as content warning labels, has measured intervention effectiveness on the typical news consumer, not the small minority of people who are most frequently exposed to misinformation. To begin filling this gap in efficacy testing, this paper describes the results of three between-subjects studies (n = 2981) that test the impact of content warning labels on people who already have extreme, radical, or misinformed views. Together, these studies provide some evidence that content labels are less effective at decreasing trust in false information for people who already have radical or misinformed views, but this result is strongly dependent on the information being intervened with. That is, content labels were more effective when they reinforced ideological positions or had no connection to prior opinions, and content labels were less effective when they undermined ideological positions or closely held misinformed viewpoints. These results reflect the well-studied theories of confirmation bias and motivated reasoning. Further, these results imply that content interventions should be studied in communities where misinformation is most frequently consumed rather than the broader population of typical news consumers.