Community Moderation and the New Epistemology of Fact Checking on Social Media
Isabelle Augenstein, Michiel Bakker, Tanmoy Chakraborty, David Corney, Emilio Ferrara, Iryna Gurevych, Scott Hale, Eduard Hovy, Heng Ji, Irene Larraz, Filippo Menczer, Preslav Nakov, Paolo Papotti, Dhruv Sahnan, Greta Warren, Giovanni Zagni · 2025
Read Community Moderation and the New Epistemology of Fact Checking on Social Media by Isabelle Augenstein, Michiel Bakker, Tanmoy Chakraborty, David Corney, Emilio Ferrara, Iryna Gurevych, Scott Hale, Eduard Hovy, Heng Ji, Irene Larraz, Filippo Menczer, Preslav Nakov, Paolo Papotti, Dhruv Sahnan, Greta Warren, Giovanni Zagni (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-20067.txt.
Introduction
Isabelle Augenstein and co-authors examine crowd fact-checking on social media in 2025. Platforms once relied on staff teams and outside checkers to flag false claims. Some large platforms now use notes written and rated by users themselves.
The authors map current approaches across major platforms and weigh the promise of crowd checks. Such notes could fight false claims at greater scale and speed. Yet views of truth are shaped by bias, politics, and culture, so crowd agreement is hard. The paper argues that crowd efforts can aid, but not replace, skilled fact-checkers.
Catalog record
| CREATOR | Isabelle Augenstein, Michiel Bakker, Tanmoy Chakraborty, David Corney, Emilio Ferrara, Iryna Gurevych, Scott Hale, Eduard Hovy, Heng Ji, Irene Larraz, Filippo Menczer, Preslav Nakov, Paolo Papotti, Dhruv Sahnan, Greta Warren, Giovanni Zagni |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 49 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2505-20067
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Community Moderation and the New Epistemology of Fact Checking on Social Media Isabelle Augenstein, Michiel Bakker, Tanmoy Chakraborty, David Corney, Emilio Ferrara, Iryna Gurevych, Scott Hale, Eduard Hovy, Heng Ji, Irene Larraz, Filippo Menczer, Preslav Nakov, Paolo Papotti, Dhruv Sahnan, Greta Warren, Giovanni Zagni Abstract Social media platforms have traditionally relied on internal moderation teams and partnerships with independent fact-checking organizations to identify and flag misleading content. Recently, however, platforms including X (formerly Twitter) and Meta have shifted towards community-driven content moderation by launching their own versions of crowd-sourced fact-checking – Community Notes. If effectively scaled and governed, such crowd-checking initiatives have the potential to combat misinformation with increased scale and speed as successfully as community-driven efforts once did with spam. Nevertheless, general content moderation, especially for misinformation, is inherently more complex. Public perceptions of truth are often shaped by personal biases, political leanings, and cultural contexts, complicating consensus on what constitutes misleading content. This suggests that community efforts, while valuable, cannot replace the indispensable role of professional fact-checkers. Here we systemically examine the current approaches to misinformation detection across major platforms, explore the emerging role of community-driven moderation, and critically evaluate both the promises and challenges of crowd-checking at scale. Introduction Social media platforms empower users to share opinions and perspectives at scale. This openness brings the persistent challenge of dealing with harmful, misleading, or otherwise objectionable content without unduly constraining freedom of expression. To navigate this tension, platforms implement content moderation policies aimed at protecting users from potentially dangerous material while preserving the integrity of public discourse. The responsibility of enforcing these rules traditionally falls upon teams of experts such as content moderators, supplemented by veracity judgments provided by third-party independent fact-checkers [ 1 , 2 , 3 ] as well as by automatic Artificial Intelligence (AI) systems working in tandem [ 4 , 5 , 6 ] . In practice, these platform enforcers determine whether to remove content, restrict its visibility, or attach warning disclaimers to a post, depending on its potential to cause harm. However, both AI-based and centralized manual moderation have limitations. AI tools for content moderation, while scalable, are marred by high rates of false positives , often over-flagging benign content and unfairly reprimanding users, and false negatives , letting truly harmful content slip by undetected [ 7 , 8 , 9 ] . Recent advances in large language models (LLMs) have shown some promise in automatic veracity prediction [ 10 , 11 ] to support fact-checkers but suffer from issues with factuality [ 12 ] and utility in practice [ 13 , 14 ] . Fact-checking experts are reliable, but cannot keep up with the relentless pace of user-generated content posted online [ 15 , 16 ] . The sheer volume and velocity of online information has strained traditional fact-checking models, leading to innovations in crowd-sourced moderation approaches such as Community Notes [ 17 ] , which aim to leverage crowd wisdom for broader and faster coverage, albeit with its own set of considerations regarding quality and bias. In an effort to keep content moderation as democratic as possible, the core idea behind this approach is to leverage the collective input of users on the platform (i.e., community ) to add context to posts that may violate content policies or contain misleading information. In this approach, platforms maintain automatic systems and internal teams in place to remove illegal or severely harmful posts containing harassment, violence, sexual exploitation, and drug-related content, while the remainder of posts are subject to community-driven moderation. In January 2025, Meta revealed a major policy change, announcing that it would end the use of third-party fact-checkers on its platforms due to alleged bias [ 18 ] . Meta stated that content that many users might view as acceptable political commentary — albeit to be consumed with a pinch of salt — may have been unnecessarily suppressed. This argument is disputed by fact-checkers, who point out that they take serious steps to maintain political impartiality, including regular internal and external reviews [ 19 ] . Neither professional fact-checking alone nor nascent community moderation systems like Community Notes offer a perfect solution to online misinformation. While professional fact-checking offers depth and rigor, its scalability is limited. Conversely, community-driven models like Community Notes promise scalability and diverse perspectives but must navigate challenges of consensus-building and potential manipulation. Additionally, it builds upon an epistemological proposition that facts are subject to consensus and negotiation, rather than objective or indisputable as traditionally intended by professional fact-checkers. Such a shift has far-reaching potential consequences for the global information eco-system more broadly. This paper explores how these distinct approaches have been deployed in content moderation, critically assesses the promises of community moderation, and highlights how collaboration between communities, experts, and technical innovations can address pervasive online misinformation. Community-Driven Content Moderation The concepts underpinning community-driven content moderation have a long history [ 20 ] . Wikipedia, for instance, has long operated on the principle of collaborative governance, relying on its users to faithfully curate and manage information for the platform [ 21 ] . Similarly, Reddit delegates some autonomy to its users, allowing a subset of users to moderate content within their respective subreddits [ 22 ] . Many social media platforms lean on user reports, not only to flag harmful content that evades content moderators, but also to help calibrate policies in response to concerns about certain types of content [ 18 , 23 ] . However, user reports that inform content moderation decisions implemented by the platforms differ from more structured approaches to community fact-checking. In the domain of community-driven content moderation on social media platforms, Twitter (now X) launched Birdwatch [ 17 ] in early 2021, the first large-scale initiative in this space, and later rebranded it Community Notes [ 24 , 25 ] . While the program’s initial iteration exposed significant shortcomings—such as a vulnerability towards targeted manipulation attempts and partisan bias affecting the notes’ writing style and approval, the company has since invested a substantial amount of resources into its refinement [ 26 , 27 ] . Key improvements include a more sophisticated algorithm to calculate helpfulness of notes, which rewards notes endorsed by a diverse set of users rather than a simple majority; and eligibility criteria for contributors to ensure participation by genuine users [ 25 ] . Following X’s lead, other social media giants, such as Meta and, to a lesser extent, TikTok and Weibo, have recently pivoted in favor of a similar community-driven moderation approach over hired experts for moderation on their platforms [ 28 , 29 , 30 ] . Some of these platforms posit community notes as a one-size-fits-all solution to the limitations of fact-checker-led content moderation. Community notes have indeed shown encouraging results on several fronts where moderation assisted by third party fact-checkers is limited, such as content coverage [ 31 , 32 , 33 ] . However, most of these results and the portrayal of community-driven content moderation as the definitive solution rest on overly-optimistic assumptions about the integrity, diversity, and efficiency of user collaboration. Just as citizen journalism’s initial promise to democratize information has faltered and been critiqued for its issues with capacity, reliability and lack