Examining accuracy-prompt efficacy in combination with using colored borders to differentiate news and social content online

Venya Bhardwaj; Cameron Martel; David G. Rand; Vicky Gabriel

In brief

Accuracy prompts reliably increase the quality of news shared on social media by reducing false-news sharing by 3.5–8.8%. Adding purple colored borders around news posts increased overall engagement with news (true and false) while decreasing engagement with non-news posts, but did not enhance the effectiveness of accuracy prompts. The study advances understanding of interventions in realistic mixed-content newsfeeds with multiple engagement options (sharing and liking).

Key claims

Accuracy prompts work in realistic newsfeeds: Accuracy prompts increased sharing discernment (true relative to false news) by ~1.7 percentage points in a simulated scrollable newsfeed environment with mixed content (news and social posts), consistent with prior lab work.
Reduced false sharing, not increased true sharing: The accuracy-prompt-only condition achieved this improvement primarily by decreasing false-news sharing 8.8% relative to control. The accuracy-prompt-plus-borders condition worked partly through increased true-news sharing.
Colored borders don't enhance accuracy prompts: Adding purple borders around news posts did not increase the efficacy of accuracy prompts compared to accuracy prompts alone. Both conditions showed similar improvements in sharing discernment.
Borders increase news engagement uniformly: Colored borders increased engagement (sharing and liking) with news posts regardless of veracity. This suggests borders work by drawing attention to news content, not by reinforcing accuracy considerations.
Engagement patterns vary by content type: False news headlines were shared most frequently, followed by true news, political social posts, and non-political social posts. The opposite pattern held for liking, with non-political social posts liked most and false news liked least.
Accuracy prompts don't harm social-post engagement: Accuracy prompts did not affect sharing of social (non-news) posts, addressing concerns that accuracy interventions would reduce platform engagement overall.

Contributions

Advances accuracy-prompt research by testing in more realistic conditions: scrollable newsfeed, mixed news and social content, multiple engagement options
Tests a novel visual differentiation intervention (colored borders) and its interaction with accuracy prompts
Provides evidence that accuracy prompts specifically target news-sharing decisions while leaving other engagement forms unaffected
Documents distinct engagement norms for different content types (news vs. social, true vs. false)

Method

Online survey experiment with 1,524 U.S. social media users (recruited via Lucid, quota-sampled for age, gender, race, region). Participants randomly assigned to control, accuracy-prompt-only, or accuracy-prompt-plus-borders conditions. Accuracy prompt: single pre-treatment task asking participants to rate accuracy of one non-political news headline (1 = "Extremely inaccurate" to 6 = "Extremely accurate"). Main task: 72-post scrollable newsfeed with 24 news posts (12 true, 12 false; balanced pro-Democratic, pro-Republican, COVID-related, vaccine-related) and 48 social posts (24 political, 24 non-political). Participants could share, like, or scroll past each post. Outcomes measured: sharing and liking rates by post type and veracity.

Results

For accuracy-prompt-only condition:

Increased sharing of true relative to false news by 1.7 percentage points (b = .017, p = .05)
Decreased false-news sharing 8.8% relative to control
No effect on liking discernment (b = -.001, p = .865)
No effect on sharing of social posts (non-news)
Effect on overall engagement (share or like) for true vs. false news (b = .017, p = .037)

For colored-borders condition:

Increased sharing of true relative to false news by 1.8 percentage points (b = .018, p = .001)
Added borders did not significantly enhance accuracy-prompt effectiveness (ps > .578)
Borders increased engagement with all news posts regardless of veracity (ps < .011)
Borders decreased liking and overall engagement with social posts (ps < .001)

Notes

Methodologically strong study with large sample, pre-registration, and attention to ecological validity (scrollable feed, mixed content, multiple engagement options). The distinction between different engagement types (sharing vs. liking) is important and underexamined in the literature. The finding that colored borders don't enhance accuracy prompts is valuable — it constrains theories about how visual design interacts with behavioral nudges. Main limitation is reliance on stated intentions in a survey setting rather than actual platform sharing behavior, though the authors cite evidence for correlation between the two. The analysis of distinct engagement patterns by content type is largely exploratory but reveals important descriptive facts about news vs. social-post engagement norms.