External
post hoc fact-checking by independent organizations began in the United States in the early 2000s. A 2018 paper found little overlap in the statements checked by different fact-checking organizations. This paper compared 1,178 published fact-checks from PolitiFact with 325 fact-checks from
The Washington Posts Fact Checker, and found only 77 statements (about 5%) that both organizations checked. For example, some are more likely to fact-check a statement about climate change being real, and others are more likely to fact-check a statement about climate change being fake.
Correcting misperceptions Studies have shown that fact-checking can affect citizens' belief in the accuracy of claims made in political advertisement. A 2020 study by
Paris School of Economics and
Sciences Po economists found that falsehoods by
Marine Le Pen during the 2017 French presidential election campaign (i) successfully persuaded voters, (ii) lost their persuasiveness when fact-checked, and (iii) did not reduce voters' political support for Le Pen when her claims were fact-checked. A 2017 study in the
Journal of Politics found that "individuals consistently update political beliefs in the appropriate direction, even on facts that have clear implications for political party reputations, though they do so cautiously and with some bias... Interestingly, those who identify with one of the political parties are no more biased or cautious than pure independents in their learning, conditional on initial beliefs." A study by
Yale University cognitive scientists
Gordon Pennycook, Adam Bear, Evan Collins, and
David G. Rand found that
Facebook tags of fake articles "did significantly reduce their perceived accuracy relative to a control without
tags, but only modestly". A Dartmouth study led by Brendan Nyhan found that Facebook tags had a greater impact than the Yale study found. A "disputed" tag on a false
headline reduced the number of respondents who considered the headline accurate from 29% to 19%, whereas a "rated false" tag pushed the number down to 16%. The Yale study found evidence of a backfire effect among Trump supporters younger than 26 years whereby the presence of both untagged and tagged fake articles made the untagged fake articles appear more accurate. Based on the findings of a 2017 study in the journal
Psychological Science, the most effective ways to reduce misinformation through corrections is by: • limiting detailed descriptions of / or arguments in favor of the misinformation; • walking through the reasons why a piece of misinformation is false rather than just labelling it false; • presenting new and credible information which allows readers to update their knowledge of events and understand why they developed an inaccurate understanding in the first place; • using video, as videos appear to be more effective than text at increasing attention and reducing confusion, making videos more effective at correcting misperception than text. Large studies by Ethan Porter and Thomas J. Wood found that misinformation propagated by
Donald Trump was more difficult to dispel with the same techniques, and generated the following recommendations: • Highly credible sources are the most effective, especially those which surprisingly report facts against their own perceived bias • Reframing the issue by adding context can be more effective than simply labeling it as incorrect or unproven. • Challenging readers' identity or worldview reduces effectiveness. • Fact-checking immediately is more effective, before false ideas have spread widely. A 2019 meta-analysis of research into the effects of fact-checking on misinformation found that fact-checking has substantial positive impacts on political beliefs, but that this impact weakened when fact-checkers used "truth scales", refuted only parts of a claim and when they fact-checked campaign-related statements. Individuals' preexisting beliefs, ideology, and knowledge affected to what extent the fact-checking had an impact. A 2019 study in the
Journal of Experimental Political Science found "strong evidence that citizens are willing to accept corrections to fake news, regardless of their ideology and the content of the fake stories." A 2018 study found that
Republicans were more likely to correct their false information on
voter fraud if the correction came from
Breitbart News rather than a non-partisan neutral source such as
PolitiFact. A 2022 study found that individuals exposed to a fact-check of a false statement by a far-right politician were less likely to share the false statement. Some studies have found that exposure to fact-checks had durable effects on reducing misperceptions, whereas other studies have found no effects. Scholars have debated whether fact-checking could lead to a "
backfire effect" whereby correcting false information may make partisan individuals cling more strongly to their views. One study found evidence of such a "
backfire effect", but several other studies did not.
Political discourse A 2015 experimental study found that fact-checking can encourage politicians to not spread
misinformation. The study found that it might help improve political discourse by increasing the reputational costs or risks of spreading
misinformation for political elites. The researchers sent, "a series of letters about the risks to their reputation and electoral security if they were caught making questionable statements. The legislators who were sent these letters were substantially less likely to receive a negative fact-checking rating or to have their accuracy questioned publicly, suggesting that fact-checking can reduce inaccuracy when it poses a salient threat." Fact-checking may also encourage some politicians to engage in "strategic ambiguity" in their statements, which "may impede the fact-checking movement's goals." A study of Trump supporters during the
2016 presidential campaign found that while fact-checks of false claims made by Trump reduced his supporters' belief in the false claims in question, the corrections did not alter their attitudes towards Trump. A 2019 study found that "summary fact-checking", where the fact-checker summarizes how many false statements a politician has made, has a greater impact on reducing support for a politician than fact-checking of individual statements made by the politician.
Informal fact-checking Individual readers perform some types of fact-checking, such as comparing claims in one news story against claims in another. Rabbi Moshe Benovitz, has observed that: "modern students use their wireless worlds to augment skepticism and to reject dogma." He says this has positive implications for values development: According to
Queen's University Belfast researcher Jennifer Rose, because fake news is created with the intention of misleading readers, online news consumers who attempt to fact-check the articles they read may incorrectly conclude that a fake news article is legitimate. Rose states, "A diligent online news consumer is likely at a pervasive risk of inferring truth from
false premises" and suggests that fact-checking alone is not enough to reduce fake news consumption. Despite this, Rose asserts that fact-checking "ought to remain on educational agendas to help combat fake news".
Detecting fake news The term
fake news became popularized with the 2016 United States presidential election, causing concern among some that online media platforms were especially susceptible to disseminating disinformation and misinformation. The language, specifically, is typically more inflammatory in fake news than real articles, in part because the purpose is to confuse and generate clicks. Furthermore, modeling techniques such as
n-gram encodings and
bag of words have served as other linguistic techniques to estimate the legitimacy of a news source. On top of that, researchers have determined that visual-based cues also play a factor in categorizing an article, specifically some features can be designed to assess if a picture was legitimate and provides us more clarity on the news. There is also many social context features that can play a role, as well as the model of spreading the news. Websites such as "
Snopes" try to detect this information manually, while certain universities are trying to build mathematical models to assist in this work. As such, researchers are calling for more work to be done regarding fake news as characterized against psychology and social theories and adapting existing data mining algorithms to apply to social media networks. Digital tools and services commonly used by fact-checkers include, but are not limited to: •
Reverse image search engines (
Google Images,
TinEye,,
Yandex Image Search Archive.today,) •
Web analytics platforms (
Similarweb) • Image and video analysis tools (
InVID, DomainBigData,) • Web mapping platforms (
Google Maps,
Google Street View,
TweetDeck,
BuzzSumo The researchers found that social media sites, Facebook in particular, to be powerful platforms to spread certain fake news to targeted groups to appeal to their sentiments during the 2016 presidential race. Additionally, researchers from
Stanford,
NYU, and
NBER found evidence to show how engagement with fake news on Facebook and Twitter was high throughout 2016. Recently, a lot of work has gone into helping detect and identify fake news through
machine learning and artificial intelligence. In 2018, researchers at
MIT's CSAIL created and tested a machine learning algorithm to identify false information by looking for common patterns, words, and symbols that typically appear in fake news. More so, they released an open-source data set with a large catalog of historical news sources with their veracity scores to encourage other researchers to explore and develop new methods and technologies for detecting fake news. In 2022, researchers have also demonstrated the feasibility of falsity scores for popular and official figures by developing such for over 800 contemporary
elites on
Twitter as well as associated exposure scores. There are also demonstrations of platform-built-in (by-design) as well
browser-integrated (currently in the form of
addons)
misinformation mitigation. Efforts such as providing and viewing structured accuracy assessments on posts "are not currently supported by the platforms". The holiday was officially created in 2016 and first celebrated on April 2, 2017. The idea for International Fact-Checking day rose out of the many
misinformation campaigns found on the internet, particularly social media sites. It rose in importance after the 2016 elections, which brought fake news, as well as accusations of it, to the forefront of media issues. The holiday is held on April 2 because "April 1 is a day for fools. April 2 is a day for facts." Activities for International Fact-Checking Day consist of various media organizations contributing to fact-checking resources, articles, and lessons for students and the general public to learn more about how to identify fake news and stop the spread of misinformation. 2020's International Fact-Checking Day focused specifically on how to accurately identify information about
COVID-19.
Limitations and controversies Research has shown that fact-checking has limits, and can even backfire, which is when a correction increases the belief in the misconception. One reason is that it can be interpreted as an
argument from authority, leading to resistance and hardening beliefs, "because identity and cultural positions cannot be disproved." In other words "while news articles can be fact-checked, personal beliefs cannot." Critics argue that political fact-checking is increasingly used as
opinion journalism. Criticism has included that fact-checking organizations in themselves are biased or that it is impossible to apply absolute terms such as "true" or "false" to inherently debatable claims. In September 2016, a
Rasmussen Reports national telephone and online survey found that "just 29% of all Likely U.S. Voters trust media fact-checking of candidates' comments. Sixty-two percent (62%) believe instead that news organizations skew the facts to help candidates they support." A paper by Andrew Guess (of Princeton University),
Brendan Nyhan (Dartmouth College) and Jason Reifler (University of Exeter) found that consumers of fake news tended to have less favorable views of fact-checking, in particular Trump supporters. The paper found that fake news consumers rarely encountered fact-checks: "only about half of the Americans who visited a
fake news website during the study period also saw any fact-check from one of the dedicated fact-checking website (14.0%)." During the
COVID-19 pandemic,
Facebook announced it would "remove false or debunked claims about the novel coronavirus which created a global pandemic", based on its fact-checking partners, collectively known as the
International Fact-Checking Network. In 2021, Facebook reversed its ban on posts speculating the
COVID-19 disease originated from Chinese labs, following developments in the
investigations into the origin of COVID-19, including claims by the Biden administration, and a letter by eighteen scientists in the journal
Science, saying a new investigation is needed because 'theories of accidental release from a lab and zoonotic spillover both remain viable." The policy led to an article by
The New York Post that suggested a lab leak would be plausible to be initially labeled as "false information" on the platform. This reignited debates into the notion of
scientific consensus. In an article published by the medical journal
The BMJ, journalist Laurie Clarke said "The contentious nature of these decisions is partly down to how social media platforms define the slippery concepts of
misinformation versus
disinformation. This decision relies on the idea of a scientific consensus. But some scientists say that this smothers heterogeneous opinions, problematically reinforcing a misconception that science is a monolith."
David Spiegelhalter, the
Winton Professor of the Public Understanding of Risk at
Cambridge University, argued that "behind closed doors, scientists spend the whole time arguing and deeply disagreeing on some fairly fundamental things". Clarke further argued that "The binary idea that scientific assertions are either correct or incorrect has fed into the divisiveness that has characterised the pandemic." Likewise, writing in
The Hedgehog Review in 2023,
Jonathan D. Teubner and Paul W. Gleason assert that fact-checking is ineffective against
propaganda for at least three reasons: "First, since much of what skillful propagandists say will be true on a literal level, the fact-checker will be unable to refute them. Second, no matter how well-intentioned or convincing, the fact-check will also spread the initial claims further. Third, even if the fact-checker manages to catch a few inaccuracies, the larger picture and suggestion will remain in place, and it is this suggestion that moves minds and hearts, and eventually actions." They also note the very large amount of false information that regularly spreads around the world, overwhelming the hundreds of fact-checking groups; caution that a fact-checker systemically addressing propaganda potentially compromises their objectivity; and argue that even descriptive statements are subjective, leading to conflicting points of view. As a potential step to a solution, the authors suggest the need of a "scientific community" to establish
falsifiable theories, "which in turn makes sense of the facts", noting the difficulty that this step would face in the digital media landscape of the Internet. Social media platforms –
Facebook in particular – have been accused by journalists and academics of undermining fact-checkers by providing them with little assistance; including "propagandist-linked organizations" promoting outlets that have shared false information such as
Breitbart and
The Daily Caller on
Facebook's newsfeed; and removing a fact-check about a false anti-abortion claim after receiving pressure from
Republican senators. In 2022 and 2023, many social media platforms such as Meta, YouTube and Twitter have significantly reduced resources in
Trust and safety, including fact-checking.
Twitter under Elon Musk has severely limited access by academic researchers to Twitter's API by replacing previously free access with a subscription that starts at $42,000 per month, and by denying requests for access under the
Digital Services Act. After the
2023 Reddit API changes, journalists, researchers and former Reddit moderators have expressed concerns about the spread of harmful misinformation, a relative lack of subject matter expertise from replacement mods, a vetting process of replacement mods seen as haphazard, a loss of third party tools often used for content moderation, and the difficulty for academic researchers to access Reddit data. Many fact-checkers rely heavily on social media platform partnerships for funding, technology and distributing their fact-checks. Commentators have also shared concerns about the use of
false equivalence as an argument in political fact-checking, citing examples from The Washington Post, The New York Times and The Associated Press where "mainstream fact-checkers appear to have attempted to manufacture false claims from progressive politicians...[out of] a desire to appear objective".
Fact-checking in countries with limited freedom of speech Operators of some fact-checking websites in China admit to
self-censorship. Fact-checking websites in China often avoid commenting on political, economic, and other current affairs. Several Chinese fact-checking websites have been criticized for lack of transparency with regard to their methodology and sources, and for following
Chinese propaganda. == Pre-publication fact-checking ==