Emphasizing publishers does not effectively reduce susceptibility to misinformation on social media

Contributions

• Empirical evidence that a platform intervention to highlight publisher information does not improve users' ability to distinguish accurate from inaccurate content
• Demonstration that publisher effects are moderated by the alignment between headline plausibility and publisher credibility
• Finding that source-based interventions may have unintended negative consequences in some cases
• Evidence that most users rely primarily on headline plausibility rather than source credibility for accuracy judgments

Method

The paper reports five studies with varying sample sizes and designs:

Study 1 (N=2,217): Participants from Amazon Mechanical Turk rated 24 headlines in Facebook News Feed format, with publisher information either absent, shown as a baseline grey-text domain, or highlighted with a logo banner. Headlines were balanced between true and false, and between sympathetic to Democrats vs. Republicans.

Studies 2 (N=1,845): Replication of Study 1 with larger sample to increase statistical power, examining potential moderators like participant partisanship and headline slant.

Study 3 (N=250): Trust ratings for 60 sources (mainstream, hyperpartisan, and fake-news outlets) to measure perceived trustworthiness. Tested the relationship between source trust and headline plausibility in predicting accuracy ratings.

Studies 4 & 5: Larger, more demographically representative samples (N=251 from MTurk, N=253 from Lucid) rating accuracy of a broader set of 600 headlines from 60 diverse sources, designed to replicate and extend the trust-plausibility relationship.

All studies used survey methodology asking participants to rate headline accuracy and sharing intentions.

Results

Finding 1: Source visibility (publisher banner vs. baseline vs. no source information) had no significant effect on accuracy judgments (F=1.59, p=.20) or sharing intentions (F=2.68, p=.068) across Studies 1 and 2. Emphasizing publisher information did not meaningfully improve truth discernment.

Finding 2: Strong positive correlation (r=.73) between trust in a given outlet and perceived plausibility of headlines from that outlet (Study 3). When no source information was provided, users relied on headline plausibility to judge accuracy, and this plausibility correlated strongly with the source's trustworthiness.

Finding 3: Publisher information only affected accuracy ratings when headline plausibility and publisher trust were mismatched. For implausible headlines from trusted sources, emphasizing the publisher increased accuracy ratings; for plausible headlines from distrusted sources, it decreased accuracy ratings. For aligned cases (plausible from trusted, implausible from distrusted), source information had minimal effect.

Finding 4: A concerning unintended consequence emerged: fact-checking 30% of plausible headlines from distrusted sources in the corpus found them to be actually true. Providing publisher information for these plausible-but-true headlines from distrusted sources increased the chance they would be mistakenly rated as false, potentially raising negative consequences from source-based interventions.

Notes

This is a rigorous, preregistered study with multiple replications that challenges an intuitive platform intervention. The finding that emphasizing publishers does not help, and may even hurt in some cases, underscores the importance of testing interventions empirically rather than relying on design intuition. The correlation between headline plausibility and source trust is noteworthy: users appear to use implicit signals about source credibility (gleaned from headline tone, topics, etc.) rather than explicit source labels to judge accuracy. The paper raises important questions about how platforms should present source information and whether decontextualizing headlines from their sources might paradoxically improve discernment.