Moral Language in Interventions Against Misinformation on Social Media

Overview

Preventing the spread of misinformation on social media is not only about providing corrections for false claims, but also about ensuring that the correct information is presented in a clear, concise, and neutral manner. The linguistic cues that fact-checkers deploy in their messaging affect the effectiveness of their interventions against misinformation. X (formerly Twitter) recently expanded its Community Notes program, which enables any user registered on the site to edit, vote on, and review potential annotations on false posts. Because Community Notes is crowdsourced, the same group of users who post on the website also write the interventions against misinformation. This method raises questions about accuracy and quality; other social media platforms employ trained fact-checkers instead of laypeople. As a metric to evaluate the quality of annotations, this study investigates the prevalence of moral language in X's crowdsourced approach to fact-checking. To do so, we collected two thousand posts and five thousand annotations and quantified the use of moral language using Moral Foundations Dictionary 2.0. Our results indicate that users utilize a significantly different moral vocabulary when posting normally versus writing an annotation. Importantly, they use less value-laden language when writing annotations than when writing posts, which indicates an effort at neutrality in their fact-checking. Given the connections between moral language and information processing, social media platforms and individual fact-checkers need to pay close attention to whether their annotations contain moral language and study the subsequent effects on belief formation.