Gordon PennycookContributions
- Extends prior work on fake news and analytic thinking by examining both fake news and hyperpartisan news (misleading but event-based coverage with strong partisan bias).
- Disentangles accuracy judgments from sharing intentions by randomly assigning participants to rate either headline accuracy or willingness to share, avoiding the confounding effect of accuracy-questioning on sharing decisions.
- Tests the predictions of two theoretical accounts of reasoning and misinformation susceptibility—classical reasoning vs. motivated system 2 reasoning—in the context of politically slanted content.
- Examines both accuracy discernment and sharing behavior across two demographically different samples (MTurk and Lucid), testing robustness.
Method
Two studies recruited approximately 1,000 American participants each using Amazon Mechanical Turk (study 1) and Lucid (study 2, quota-sampled to match national demographics on age, gender, ethnicity, region). Participants rated 30 headlines (10 false, 10 hyperpartisan, 10 true from mainstream sources) in Facebook post format. Headlines were either pro-Democrat or pro-Republican. Participants were randomly assigned to one of two conditions: accuracy condition (rate whether headline describes an event that happened in an accurate and unbiased way) or sharing condition (indicate willingness to share on social media). Analytic thinking was measured using the Cognitive Reflection Test (CRT), a seven-item task that assesses deliberative reasoning. Participants also completed measures of political partisanship, ideology, and social media use.
Results
Accuracy judgments: CRT was not positively associated with perceiving politically consistent false or hyperpartisan headlines as less accurate, contrary to the motivated reasoning account. Instead, CRT was consistently positively associated with better discrimination between true and false headlines (true-false discernment) and between true and hyperpartisan headlines (true-hyperpartisan discernment), across both samples and political slants. This supports the classical reasoning account: people higher in analytic thinking are generally better at distinguishing high-quality from low-quality news, regardless of political alignment.
Willingness to share: CRT was not positively associated with increased willingness to share false or hyperpartisan news. Rather, CRT was often negatively associated with willingness to share false and hyperpartisan content (especially in the MTurk sample), and was associated with improved media truth discernment for sharing decisions.
Sample differences: The Lucid sample (nationally representative) showed weaker associations between CRT and accuracy/sharing measures than the MTurk sample. This may reflect lower variance in CRT scores in the representative sample or other demographic factors.
Notes
The paper carefully separates accuracy judgments from sharing intentions, a methodological strength over prior work that assessed both sequentially and found that questioning accuracy itself changes sharing behavior. The finding that analytic thinking supports accurate belief formation rather than political motivated reasoning is important for understanding the psychological pathways to misinformation resistance. The weaker effects in the representative Lucid sample suggest population-level generalization remains an open question, and differences in CRT variance across samples warrant further investigation. The focus on hyperpartisan (rather than blatantly false) headlines makes this a particularly relevant test case for real-world misinformation where falsity is often implicit rather than explicit.