Exposure to Social Engagement Metrics Increases Vulnerability to Misinformation

• The study asks whether exposure to social engagement metrics affects people’s propensity to share content.Source: Research questions · p. 1
• It asks whether high engagement increases liking or sharing of questionable content.Source: Research questions · p. 1
• It also asks whether high engagement makes users less likely to fact-check low-credibility sources.Source: Research questions · p. 1

• Fakey simulates a social media feed and presents current articles from mainstream and low-credibility sources with randomly generated engagement metrics.Source: Essay summary · p. 1
• The 19-month deployment included over 8,500 users and approximately 120,000 articles, half from low-credibility sources.Source: Essay summary · p. 1
• Higher displayed engagement was associated with greater propensity to share questionable content and less propensity to fact-check it.Source: Essay summary · p. 1
• The authors call for platforms to reconsider how engagement metrics are displayed and for further research on designs that avoid facilitating misinformation or hindering legitimate information.Source: Essay summary · p. 1
• Figure 1 depicts a news post in the simulated social media feed used by the game.Source: Figure 1 · p. 2

• Existing responses include technical detection approaches and interface credibility indicators, which can reduce sharing of flagged content.Source: Implications · p. 2
• Exposure to high engagement metrics is associated with less scrutiny of low-credibility content and greater likelihood of liking or sharing it.The authors interpret metrics as signals that can mimic repeated exposure and perceived validation by other users.Source: Implications · pp. 2–3
• Prior work has examined environmental, emotional, and individual factors in online contagion, while empirical research on social-feed elements has been limited.Source: Implications · p. 2
• Engagement metrics may be vulnerable to manipulation because users can mistake boosted counts for independent user activity.The authors note that disinformation campaigns use inauthentic accounts to create an initial appearance of virality.Source: Implications · p. 3
• The authors call for research on alternative metric displays that reduce misinformation sharing without reducing sharing of legitimate, reliable content.Source credibility and topic sensitivity are proposed as properties that could inform differential display.Source: Implications · p. 3
• The authors recommend investigating literacy campaigns that teach users to prioritize source trustworthiness over engagement signals.They also discuss combining literacy efforts with pauses, limits on automated or high-speed sharing, and other approaches.Source: Implications · p. 3

Finding 1: High levels of social engagement results in lower fact checking and higher liking/sharing, especially for low-credibility content.

• For low-credibility articles, social engagement correlated with liking/sharing (Spearman ρ = 0.97, p < 0.001) and fact checking (ρ = −0.97, p < 0.001).For mainstream articles, the corresponding correlations were weaker (ρ = 0.66 for liking/sharing; ρ = −0.62 for fact checking).Source: Finding 1: High levels of social… · pp. 3–4

Finding 2: People are more vulnerable to low-credibility content with high social engagement.

• Figure 2 shows individual-level mean rates; for low-credibility articles, liking/sharing and fact-checking rates differed across all engagement-bin pairings (p < 0.001).Kruskal-Wallis tests found differences across bins for liking/sharing (χ2(2) = 417.14) and fact checking (χ2(2) = 214.26), both p < 0.001.Source: Figure 2 · p. 4, Finding 2: People are more vulnerable… · p. 4
• For mainstream articles, engagement level also significantly affected liking/sharing (χ2(2) = 161.80, p < 0.001) and fact checking (χ2(2) = 576.37, p < 0.001) across bins.Source: Finding 2: People are more vulnerable… · p. 4

Social media simulation

• Fakey 4 was an online news-literacy game simulating fact-checking in a Facebook- or Twitter-like feed, presenting batches of ten articles with social-engagement metrics and buttons to share, like, fact-check, or skip.The single metric combined shares and likes; values were randomly drawn from an approximately log-normal distribution and were not drawn from empirical data.Source: Social media simulation · p. 5

Content selection

• Articles came from mainstream and low-credibility domains; five from each source type were randomly chosen per round, without topic filtering.Mainstream sources were manually selected; low-credibility sources were flagged by news and fact-checking organizations.Source: Content selection · p. 5

Data collection

• A 19-month deployment yielded about 8,606 users and 120,000 articles; the study collected no demographics, and the mobile app was available in four English-speaking countries.Participants originated mostly from the United States (78%), followed by Australia (8%), the UK (4%), Canada (3%), Germany (3%), and Bulgaria (2%).Source: Data collection · pp. 5–6

Limitations

• The simulated game differs from real platforms and captures only Like/Share; equal source representation and misinformation priming may raise fact-checking rates, though authors say these effects should not alter engagement-signal results.User backgrounds and content types may offer further insight into vulnerability but were not examined.Source: Limitations · p. 6

Funding

• M.A. and F.M. were partly supported by the Democracy Fund, Craig Newmark Philanthropies, and Knight Foundation; funders had no role in the study.Source: Funding · p. 9