Adam Berinsky
Gordon Pennycook
David RandKey claims
- The accuracy-sharing gap: Participants showed fair discernment when asked to judge headline accuracy (true headlines: 67.7% rated accurate, false: 40.9%), but this discernment vanished when deciding whether to share (true: 54.1%, false: 48.3% — no significant difference).
- Effectiveness of prompts: Three treatments significantly increased sharing discernment by roughly 50%: (1) asking participants to evaluate a single non-COVID headline's accuracy; (2) providing four digital literacy tips; (3) asking about the importance of sharing only accurate news. A longer evaluation task with feedback increased discernment by ~100%.
- Ineffective approach: Descriptive norms (informing users that others think accuracy is important) did not increase sharing discernment on its own.
- Mechanism: Treatments work primarily by shifting attention to accuracy, not through knowledge transfer or commitment. The effect is stronger for participants who are more attentive, analytical, engaged with news, and college-educated.
- Broad generalizability: Treatment effects did not vary significantly by gender, race, partisanship, or COVID-19 concern level, though effects were strongest for middle-aged participants. Even less analytical, less news-engaged, and non-college-educated participants showed significant treatment effects.
Notes
A well-designed empirical study with large sample and careful experimental controls. The main contribution is demonstrating that multiple lightweight approaches to prompting accuracy work and generalize broadly, giving practitioners a "menu" of interventions. The distinction between accuracy judgments and sharing intentions is striking and important. Findings are limited to stated sharing intentions in a survey setting rather than actual sharing behavior, though prior work suggests reasonable correlation between the two. The study is published in a peer-reviewed outlet (HKS Misinformation Review) with open materials and data.