Accurate text generating inaccurate beliefs: a study on the effect of publishing beach pictures alongside articles about COVID-19
Preprint
Jacob Derechin*, Soham Sankaran*, Nicholas A. Christakis
Abstract: In a 3021 person randomized controlled experiment measuring the impact of publishing pictures of beaches alongside articles about the spread of COVID-19 on reader beliefs, we found that putting beach pictures alongside articles of this nature increased reader perceptions of the risk associated with contracting COVID-19 at the beach relative to restaurants, and also increased the proportion of people who rank the beach more risky than restaurants. The change in perception occurred despite the fact that the text of those articles was either neutral on the subject of relative risk or explicitly noted that outdoor spaces such as beaches are less fertile ground for COVID-19 transmission than indoor spaces like restaurants. We use this result to argue for the adoption of a new standard for journalistic accuracy based on the empirically measured beliefs produced by a given media artifact in readers, as opposed to merely the technical accuracy of the piece in the eyes of an expert.