Analysis | Twitter has an experimental crowdsourced fact-checking feature – The Washington Post
Sign in
A newsletter briefing on the intersection of technology and politics.
with research by Aaron Schaffer
A newsletter briefing on the intersection of technology and politics.
Good morning! I’m Will Oremus, filling in for your regular host Cristiano Lima today. You can reach me at [email protected]
Below: A former Twitter executive has filed an explosive whistleblower complaint, and lawmakers release a revamped bill to let news outlets negotiate with major tech firms. First:
As the big social media platforms trot out familiar misinformation policies ahead of the 2022 midterm elections, Twitter is holding what could be a wild card. But it’s one the company is reluctant to play before it’s sure it has a winning hand.
Twitter’s experimental, crowdsourced fact-checking project, called Birdwatch, is quietly gaining momentum behind the scenes, and the company is gradually beginning to add new features and show it to more users, Twitter vice president of product Keith Coleman told The Technology 202. Internal research finds it is leading ordinary users to share less misinformation on the site, Coleman said.
Birdwatch’s apparent progress represents a potential bright spot for a company whose struggles to police misinformation have drawn fire from both parties.
The project, which lets ordinary Twitter users write fact-checking “notes” on others’ tweets and then rate one another’s notes as “helpful” or “not helpful,” has been brewing as a small beta test for 18 months, largely out of the public eye. After a year in which Birdwatch notes were visible only to a small group of volunteer participants, Twitter began showing some of the highest-rated notes to a test group of ordinary users in March.
Now it’s showing them to more people, encouraged by findings that the Birdwatch notes rated highly by users tend to be broadly accurate and informative, as measured by independent fact-checkers, said Coleman, who leads a team of about 15 Twitter employees working on the project. In one test, users who saw a misleading tweet with a Birdwatch note attached were 20 to 40 percent less likely to agree with the tweet’s substance than those who saw only the tweet without the Birdwatch note.
An analysis of public Birdwatch data by my colleague Jeremy Merrill confirms that Birdwatch has picked up steam. While the number of participants writing notes hasn’t noticeably increased, the notes they do write are receiving significantly more ratings in recent months, suggesting that more ordinary Twitter users are seeing them and interacting with them. In June and July, Birdwatch notes received more than 1,000 ratings per day, on average — more than double the rate for the first three months of the year.
But anyone expecting Birdwatch to play a significant role in the upcoming midterms is likely to be disappointed.
Coleman said the project continues to move on its own, intentionally cautious timetable, and won’t be rushed out for any particular political cycle. For now, the pool of users who can write their own Birdwatch notes remains limited to the thousands of volunteers who applied and were accepted into the pilot program, while the fraction of all Twitter users who can view those notes remains small. And the project remains U.S.-only for the foreseeable future.
Crucially, the Birdwatch labels appear to affect not only users’ perceptions, but their behavior.
Users who see a highly rated Birdwatch note on a tweet are less likely to share the tweet to their own followers, Coleman said. If that holds true as the project rolls out more widely, it would mean that user-written fact-checking labels have the potential to reduce the spread of misinformation without Twitter itself having to take any action.
That’s noteworthy at a time when Twitter and other social media giants face criticism — and state legislation — from Republican lawmakers who see the platforms’ top-down efforts to limit misinformation as a form of censorship.
“The fact that it’s so human-driven, versus driven by one organization, is really powerful and really exciting, and it’s different than the approaches that have historically been taken,” Coleman said.
To guard against trolls gaming the system, Birdwatch’s algorithms assign more weight to notes that are rated as “helpful” by people from across the partisan spectrum. Coleman said the team is also working to ensure Birdwatch users with a strong reputation for helpful contributions have more influence than those without such a track record.
So far, Birdwatch users appear to be motivated as much by politics as by truth, said David Rand, an MIT professor who researches misinformation and serves as an unpaid adviser on the Birdwatch project. Misleading tweets about hot-button political issues and covid get far more notes than other forms of misinformation, such as non-covid-related health misinformation. And Rand said participants are much more likely to rate notes “helpful” from members of their own party.
Yet that adversarial element doesn’t mean the project is doomed. As long as most participants focus on tweets that really are misleading, having each side scrutinize the other’s claims seems to produce generally reliable results in most cases, Rand said.
Still, there are already signs that the increased rollout of Birdwatch is likely to bring some confusion and controversy.
For instance, Birdwatch notes on August tweets by Joe Biden and White House press secretary Karine Jean-Pierre about the inflation rate drew criticism from some media figures who appeared to blame Twitter itself for notes they found misguided.
Coleman acknowledged that trust in Birdwatch could be undermined if controversial or questionable notes start appearing widely in users’ feeds. He said that’s why the team is moving so slowly.
In an explosive whistleblower complaint, former Twitter head of security Peiter Zatko accused Twitter of deceiving its board and federal regulators about “extreme, egregious deficiencies” in its defenses against hackers and meager efforts to fight spam, Joseph Menn, Elizabeth Dwoskin and Cat Zakrzewski report. One of the most serious accusations in the complaint — which was sent to the Securities and Exchange Commission, Federal Trade Commission and Department of Justice — argues that Twitter violated an 11-year-old FTC settlement by claiming that it had a solid security plan.
By age 30, Zatko wrote a powerful password-cracking tool, testified to Congress under hacker handle “Mudge,” and co-founded one of the first hacking consultancies backed by venture capital, Joseph reports. Before Twitter’s then-chief executive Jack Dorsey recruited Zatko, he had worked at the Defense Advanced Research Projects Agency, Motorola Mobility, Google and Stripe.
Twitter described the complaint as an opportunistic attack. “Security and privacy have long been top companywide priorities at Twitter,” said Rebecca Hahn, Twitter’s global vice president of communications. “Mr. Zatko was fired from Twitter more than six months ago for poor performance and leadership, and he now appears to be opportunistically seeking to inflict harm on Twitter, its customers, and its shareholders.” The company’s security practices are within industry standards and the company has rules about who can access its systems, Hahn said.
The complaint could give Elon Musk ammunition in his bid to exit his $44 billion deal to purchase Twitter, Faiz Siddiqui and Elizabeth Dwoskin report. Musk has argued that Twitter’s estimate that bots and spam accounts amount to less than 5 percent of its users is false, and he terminated his agreement to buy Twitter by arguing that the bot miscount amounts to a “material adverse effect.” Musk also countersued Twitter for allegedly misleading his team. He also accused the company of fraud and breach of contract.
“Musk is correct,” Zatko’s complaint alleges, according to a copy obtained from a senior Democratic aide on Capitol Hill. “Twitter executives have little or no personal incentive to accurately ‘detect’ or measure the prevalence of spam bots.”
A former high-level Twitter executive could be a star witness for Musk’s team, legal experts told The Post. Musk would be able to subpoena a former employee with relevant information, they said. Zatko’s attorneys, from nonprofit law firm Whistleblower Aid, said they hadn’t interacted with Musk’s team; however, they said that Zatko would respond to subpoenas.
“Musk’s legal team still faces a steep climb in demonstrating that any of the allegations by Zatko — even if they are proved — so drastically change the value of Twitter that he should be able to pull out of the deal,” my colleagues write. “Legal experts say the court where the trial is being held is traditionally business-friendly, and a declaration of a material adverse effect is rare.”
Twitter stands by its SEC filings, methodologies for counting users and statements about the percentage of spam accounts, Hahn said.
A revamped version of the bipartisan legislation expands the amount of time publishers have to negotiate, has new requirements for good-faith negotiations and revises the definition of an “online platform,” Bloomberg Government’s Maria Curi reports. The authors of the bill also aimed to assuage concerns about the copyright implications of the bill, which were criticized by groups like the Electronic Frontier Foundation.
The legislation seeks “to protect small newspapers that face diminishing subscriptions and advertisement revenue,” Curi writes. “The bipartisan measure, first introduced in March 2021, was on the schedule for a Senate Judiciary Committee markup earlier this month but the panel postponed its consideration.”
An updated rendering of Meta chief executive Mark Zuckerberg‘s face (which Zuckerberg posted after criticism of a previous image he posted) was the subject of questions about the future of the metaverse — and more criticism. Reporter Queenie Wong:
The many faces of Mark Zuckerberg makes me wonder how the metaverse will impact body image. Anyone else notice that these avatars don’t have any wrinkles? pic.twitter.com/TrcXCRkocE
Reporter Ryan Mac:
Wow. Mark Zuckerberg basically got cyberbullied so much over his company’s metaverse graphics that he announced an update. pic.twitter.com/Eu7108qCUu
Creative director James Bareham:
Looks like Zuck is coming to Fortnite https://t.co/5QeeHD3Lgw
Musk seeks documents from Jack Dorsey in fight over Twitter deal (Reuters)
Norway wants Facebook fined for illegal data transfers (Politico Europe)
TikTok and Meta ban self-described misogynist Andrew Tate (Brittany Shammas)
Musk seeks dismissal of Twitter investor suit as ‘hypothetical’ (Bloomberg)
Amazon warehouse quietly addresses safety concerns following a worker’s sudden death (NBC News)
Tesla loses challenge to California agency suing for race bias (Reuters)
Banks, crypto lobby clash with lawmakers over Fed digital dollar (Politico)
A dad took photos of his naked toddler for the doctor. Google flagged him as a criminal. (The New York Times)
The low threshold for face recognition in New Delhi (WIRED)
How a photo of a woman yelling in a guy’s ear became a viral meme (María Luisa Paúl)
The ball indeed stopped him 💀
(via danielchacon69/TT) pic.twitter.com/VPDFwd5gBk
That’s all for today — thank you so much for joining us! Make sure to tell others to subscribe to The Technology 202 here. Get in touch with tips, feedback or greetings on Twitter or email.
