Less than an hour after President Biden endorsed Vice President Kamala Harris as the Democratic candidate for president on Sunday, users on X falsely claimed she was “ineligible” to run for office because her parents were not born in the United States.
The false assertion spread quickly. Under one post with the untrue claim about Ms. Harris, which was shared thousands of times and received nearly 137,000 views, another X user quickly added a warning that the post was wrong and that Ms. Harris was eligible to run as a U.S. citizen.
“Kamala Harris was born in Oakland, California, USA,” the note said. “This makes her a natural born citizen. Eligible to run for President of the United States.”
But on dozens of other posts containing the same misinformation, no such corrections appeared.
Elon Musk has done away with most content moderation rules on X since buying the social network in 2022. He has instead relied on a program called Community Notes, which lets a group of users write fact-checking labels and vote on whether they are helpful. Those approved are added to misleading posts by an algorithm. X has determined that the program is sufficient to police most misinformation about the presidential election in November, two people familiar with the company’s plans said.
But Community Notes has been far from consistent. Even with a corrective label attached, misleading claims can spread across X, according to misinformation researchers. And because users from across the political spectrum have to agree on how to correct posts, some comments and fact checks about divisive topics never surface at all.
Nearly 8,000 fact checks have been drafted about immigration on Community Notes, but only 471 of them have been approved by users and made public on X, according to MediaWise, a media literacy program at the Poynter Institute. Only 4 percent of Community Notes about abortion have been made visible.
X’s reliance on Community Notes matters because more social platforms are starting to adopt the same tactic to deal with misinformation. YouTube and Meta, which owns Facebook and Instagram, have cut back on content moderators, and YouTube recently began experimenting with a program to let users moderate its site.
It also means that X may be flooded by misleading claims surrounding the vote in November as the company steps back from policing content. And X’s decision to rely on an increasingly politically divided group of users could prime the platform for further misuse. #StoptheSteal posts on right-wing and mainstream platforms helped stoke an attack on the U.S. Capitol on Jan. 6, 2021. Last year, supporters of former President Jair Bolsonaro of Brazil raided federal buildings after similar lies spread online about the country’s election.
“If they truly want to rely on Community Notes to root out political misinformation, it will not work,” said Alex Mahadevan, the director of MediaWise at the Poynter Institute. “We’re so polarized that nobody can agree anymore.”
X and Mr. Musk did not respond to requests for comment. X will continue to enforce a limited set of rules against sharing misleading information about how to vote, or calling for violence near polling stations, by labeling such posts and preventing other users from sharing them, according to a company policy.
Mr. Musk has praised the program, frequently responding to posts about published Notes on X with “Community Notes ftw,” using slang that means “for the win.”
X, formerly known as Twitter, once had thousands of content moderators to police misinformation and hundreds of trust and safety employees who took down problematic posts. When Mr. Musk bought the company, he argued that the site was overly censored and laid off more than 75 percent of the employees.
Community Notes has become one of the few remaining ways to handle misinformation on X. Twitter created the program, originally called Birdwatch, in 2021. Much like the system used by Wikipedia for updating its pages, Community Notes relies on a group of approved X users from across the political spectrum who can write a comment about a misleading post and ask others in the group to rate it. If enough users from differing viewpoints rate a comment as being helpful, it is appended to the original post by the algorithm.
On Thursday, X made it possible for all users to request fact-checking with Community Notes when they notice a misleading post.
Researchers have found that the program was effective at countering misinformation about Covid vaccines. In that case, the notes “emerged as an innovative solution, pushing back with accurate and credible health information,” said John W. Ayers, the vice chief of innovation in the Division of Infectious Disease and Global Public Health at the University of California, San Diego, School of Medicine, who wrote a report in April on the topic.
But the community fact-checking program also frequently errs. When Hurricane Beryl was forming over the Atlantic Ocean on June 27, AccuWeather posted a forecast that accurately predicted its path toward the Yucatán Peninsula. But an inaccurate Community Note was attached, saying the forecast was backed by very little data and couldn’t be verified.
Although other forecasters, including the National Hurricane Center, issued similar predictions, the Community Note remained on AccuWeather’s post until July 2, when the storm had already killed several people in the Caribbean.
“The Community Note was misinformation,” said Steve Smith, AccuWeather’s chief executive. “We were very concerned about having that note up there when we were trying to provide lifesaving information.”
Mr. Musk has also publicly posted about Community Notes that he wants added, allowing him to flex his muscle as the platform’s owner and its most followed user, with more than 190 million followers.
On July 1, Mr. Musk posted a screenshot of a post by Ms. Harris claiming that former President Donald J. Trump would enact a national abortion ban. In the screenshot, a Community Note was attached citing the times that Mr. Trump has said he would not ban abortion.
“When will politicians, or at least the intern who runs their account, learn that lying on this platform doesn’t work anymore?” Mr. Musk said.
But the Community Note was later removed from Ms. Harris’s message, as fact checkers disputed its validity.
“Elon Musk can Google it: Donald Trump wants to ban abortion nationwide, allow his allies to ban medication abortion and contraception, and use the power of the government to harass and punish women,” said James Singer, who spoke this month on behalf of the Biden campaign.
After the July 13 assassination attempt on Mr. Trump, Community Notes struggled to keep up with the flood of misinformation. The word “staged” trended on X for hours. Some posts claiming Mr. Trump had faked the attack were corrected by Community Notes, while others spread unchecked.
The program worked in some cases. A news aggregation account with more than a million followers claimed that the shooter was “a Chinese man”; a Community Note on the post clarified that the Federal Bureau of Investigation had identified him as a white male. A post condemning the violence from Representative Bennie Thompson, a Democrat of Mississippi, received a Community Note that noted his support for legislation that would remove security details from former presidents with criminal convictions.
But other posts that contained misinformation didn’t receive notes. A prominent MAGA account, which had been banned by Twitter but reinstated by X under Mr. Musk, racked up 1.4 million views on a post claiming that the attempted assassination was an “INSIDE JOB.” A pop-punk vocalist accumulated 9.3 million views on a post claiming “NOTHING HAS EVER BEEN MORE STAGED.” The post also asserted that Mr. Trump was not strong enough to push his Secret Service detail away and raise his fist in the air, implying that his security detail had intentionally allowed him to pose for photos after the incident.
On Monday, after Mr. Biden’s endorsement of Ms. Harris, the hashtag #ProofOfLife trended on the platform. Many X users claimed that Mr. Biden was dying and that his signature on the letter saying he was withdrawing from the presidential race had been forged. Most lacked Community Notes.
“The system is designed to fail on political topics,” said Mr. Mahadevan, the MediaWise director.
Source: nytimes.com