Behind the black box: The moderating role of the machine heuristic on the effect of transparency information about automated journalism on hostile media bias perception

Abstract

Facing historically low levels of public trust, journalists had been increasingly interested in the potential of artificial intelligence to produce news content. Some have suggested that Automated Journalism (AJ) may reduce Hostile Media Biases (HMB), where partisans perceive balanced articles as slanted against their side. However, empirical evidence for the hypothesis remains limited and inconclusive. In this study, we examine whether the effectiveness of AJ at reducing HMB perceptions could be enhanced by disclosure of transparency information about how the algorithm works. We conducted an online experiment (N = 264 US adults) in which participants were randomly assigned to read a balanced news article about gun control written by different authors (AJ, AJ + transparency information, journalist, student, no author). Our findings indicate that AJ transparency, on average, did not significantly reduce HMB compared to AJ along. A significant interaction effect was identified: participants who strongly endorsed the machine heuristic were less likely to perceive the content in the AJ transparency condition, but not that of other conditions, as biased. Theoretical and practical implications are discussed.