Emotion may predict susceptibility to fake news but emotion regulation does not seem to help

Bence Bago; Leah Rosenzweig; Adam Berinsky; David Rand

In brief

Emotional responses predict susceptibility to false headlines: people experiencing any emotion (except anger) show diminished truth discernment. Four studies test whether emotion regulation techniques (suppression, reappraisal) improve accuracy. Despite clear correlational effects, the interventions produce no consistent improvements, suggesting that emotional responses alone do not drive misinformation belief, or that emotion regulation as typically administered is ineffective against headlines.

Contributions

Directly tests the hypothesis that emotional arousal causally drives belief in false headlines.
Demonstrates a clear trade-off in emotion effects: most emotions impair truth discernment, but anger (typically associated with intuition) improves it on partisan-discordant headlines.
Tests two well-established emotion regulation techniques (cognitive reappraisal, emotional suppression) in an intervention setting and finds neither helps.
Provides large-sample evidence (4 studies, N > 10,000 total) that trait-level emotion regulation does not correlate with better discernment.

Method

Study 1 (Correlational). 998 participants rated 16 headlines (8 true, 8 false; half partisan-concordant, half partisan-discordant) and reported emotional responses using 6 basic emotion icons (anger, disgust, fear, happiness, sadness, surprise). Measured headline accuracy judgments and sharing intentions. Used mixed-effect logistic regression with items and subjects as random intercepts.

Studies 2–4 (Interventions). 1,007 (Study 2), 3,001 (Study 3), and 3,032 (Study 4) participants randomly assigned to Control, Emotional Suppression, or Emotional Reappraisal (Study 2 only). In intervention conditions, before each headline, participants received instructions to either inhibit emotional expression or reframe the headline unemotionally. Same accuracy and sharing measures as Study 1. Studies 3–4 were higher-powered replications with slightly different headline sets.

Results

Study 1. In 69% of trials, people reported experiencing at least one of the six emotions. Emotional response significantly predicts worse accuracy (b = 2.3, p < 0.0001): people with emotions are more likely to believe both true and false headlines, with a net effect of worse truth discernment. Exception: anger correlates with increased discernment (b = 0.46, p = 0.002), particularly on discordant headlines (b = 1.44, p < 0.0001). Trait-level emotion regulation (reappraisal and suppression measured via questionnaire) shows no association with accuracy.

Study 2. Emotion suppression and reappraisal interventions show no main effect on accuracy (b = 0.06, p = 0.65). A post-hoc analysis on high-anger items suggests reappraisal may help on politically concordant false headlines (b = 0.19, p = 0.04), but this was unpreregistered and not replicated.

Studies 3–4. Emotion suppression has no main effect (Study 3: b = 0.00001, p = 0.99; Study 4: b = 0.03, p = 0.55), no significant interaction with headline veracity (Study 3: b = -0.05, p = 0.46; Study 4: b = -0.12, p = 0.16), and no three-way interaction with concordance and veracity. Pooled analysis across all intervention studies (N = 6,702) confirms no effect.

Notes

The null intervention results are striking and somewhat unexpected. The authors correctly note two caveats: (1) trait-level emotion regulation (from questionnaires) differs from state-level regulation (the intervention), and (2) the interventions may have inadvertently drawn attention to emotion and away from headline content. The positive association between anger and improved discernment on discordant headlines is theoretically interesting but (as the authors acknowledge) correlational and could reflect reverse causation — high-accuracy people may feel more justified anger at partisan falsehoods. The paper does not establish that emotion itself causes misinformation belief, only that it correlates with worse discernment in Study 1. The intervention studies suggest emotion is not a sufficient causal mechanism, but do not rule out emotion as a contributing factor. Worth noting: the work was largely conducted pre-pandemic (Studies 3–4 in 2020–2021), so generalizability to covid-era misinformation remains open.