Author Topic: Facebook Says Its Rules Apply to All 2/2  (Read 55 times)

Online javajolt

  • Administrator
  • Hero Member
  • *****
  • Posts: 35150
  • Gender: Male
  • I Do Windows
    • windows10newsinfo.com
    • Email
Facebook Says Its Rules Apply to All 2/2
« on: September 14, 2021, 04:10:03 PM »
◄ part 1


Flubbed calls

Mr. Zuckerberg has consistently framed his position on how to moderate controversial content as one of principled neutrality. “We do not want to become the arbiters of truth,” he told Congress in a hearing last year.

Facebook’s special enforcement system for VIP users arose from the fact that its human and automated content-enforcement systems regularly flub calls.

Part of the problem is resources. While Facebook has trumpeted its spending on an army of content moderators, it still isn’t capable of fully processing the torrent of user-generated content on its platforms. Even assuming adequate staffing and a higher accuracy rate, making millions of moderation decisions a day would still involve numerous high-profile calls with the potential for bad PR.

Facebook wanted a system for “reducing false positives and human workload,” according to one internal document. The XCheck system was set up to do that.

To minimize conflict with average users, the company has long kept its notifications of content removals opaque. Users often describe on Facebook, Instagram or rival platforms what they say are removal errors, often accompanied by a screenshot of the notice they receive.

Facebook pays close attention. One internal presentation about the issue last year was titled “Users Retaliating Against Facebook Actions.”

“Literally all I said was a happy birthday,” one user posted in response to a botched takedown, according to the presentation.

“Apparently Facebook doesn’t allow complaining about paint colors now?” another user complained after Facebook flagged as hate speech the declaration that “white paint colors are the worst.”

“Users like to screenshot us at our most ridiculous,” the presentation said, noting they often are outraged even when Facebook correctly applies its rules.

If getting panned by everyday users is unpleasant, inadvertently upsetting prominent ones is potentially embarrassing.

Last year, Facebook’s algorithms misinterpreted a years-old post from Hosam El Sokkari, an independent journalist who once headed the BBC’s Arabic News service, according to a September 2020 “incident review” by the company.

In the post, he condemned Osama bin Laden, but Facebook’s algorithms misinterpreted the post as supporting the terrorist, which would have violated the platform’s rules. Human reviewers erroneously concurred with the automated decision and denied Mr. El Sokkari’s appeal.

As a result, Mr. El Sokkari’s account was blocked from broadcasting a live video shortly before a scheduled public appearance. In response, he denounced Facebook on Twitter and the company’s own platform in posts that received hundreds of thousands of views.

Facebook swiftly reversed itself, but shortly afterward mistakenly took down more of Mr. El Sokkari’s posts criticizing conservative Muslim figures.

Mr. El Sokkari responded: “Facebook Arabic support team has obviously been infiltrated by extremists,” he tweeted, an assertion that prompted more scrambling inside Facebook.

After seeking input from 41 employees, Facebook said in a report about the incident that XCheck remained too often “reactive and demand-driven.” The report concluded that XCheck should be expanded further to include prominent independent journalists such as Mr. El Sokkari, to avoid future public-relations black eyes.

As XCheck mushroomed to encompass what the documents said are millions of users worldwide, reviewing all the questionable content became a fresh mountain of work.


Whitelist status

In response to what the documents describe as chronic underinvestment in moderation efforts, many teams around Facebook chose not to enforce the rules with high-profile accounts at all—the practice referred to as whitelisting. In some instances, whitelist status was granted with little record of who had granted it and why, according to the 2019 audit.

“This problem is pervasive, touching almost every area of the company,” the 2019 review states, citing the audit. It concluded that whitelists “pose numerous legal, compliance, and legitimacy risks for the company and harm to our community.”


Facebook is trying to eliminate the practice of whitelisting, the documents show. Its headquarters in Menlo Park.
PHOTO: IAN BATES FOR THE WALL STREET JOURNAL


A plan to fix the program, described in a document the following year, said that blanket exemptions and posts that were never subsequently reviewed had become the core of the program, meaning most contents from XCheck users wasn’t subject to enforcement. “We currently review less than 10% of XChecked content,” the document stated.

Mr. Stone said the company improved that ratio during 2020, though he declined to provide data.

The leeway given to prominent political accounts on misinformation, which the company in 2019 acknowledged in a limited form, baffled some employees responsible for protecting the platforms. High-profile accounts posed greater risks than regular ones, researchers noted, yet were the least policed.

“We are knowingly exposing users to misinformation that we have the processes and resources to mitigate,” said a 2019 memo by Facebook researchers, called “The Political Whitelist Contradicts Facebook’s Core Stated Principles.” Technology website The Information previously reported on the document.

In one instance, political whitelist users were sharing articles from alternative-medicine websites claiming that a Berkeley, Calif., doctor had revealed that chemotherapy doesn’t work 97% of the time. Fact-checking organizations have debunked the claims, noting that the science is misrepresented and that the doctor cited in the article died in 1978.

In an internal comment in response to the memo, Samidh Chakrabarti, an executive who headed Facebook’s Civic Team, which focuses on political and social discourse on the platform, voiced his discomfort with the exemptions.

“One of the fundamental reasons I joined FB Is that I believe in its potential to be a profoundly democratizing force that enables everyone to have an equal civic voice,” he wrote. “So having different rules on speech for different people is very troubling to me.”

Other employees said the practice was at odds with Facebook’s values.

“FB’s decision-making on content policy is influenced by political considerations,” wrote an economist in the company’s data science division.

“Separate content policy from public policy,” recommended Kaushik Iyer, then lead engineer for Facebook’s civic integrity team, in a June 2020 memo.


Source: May 2019 comment from Samidh Chakrabarti, Facebook's then-head of Civic Integrity, commenting on
an internal presentation titled "The Political Whitelist Contradicts Facebook's Core Stated Principles."


Buzzfeed previously reported on elements of these documents.

That same month, employees debated on Workplace, the internal platform, about the merits of going public with the XCheck program.

As the transparency proposal drew dozens of “like” and “love” emojis from colleagues, the Civic Team’s Mr. Chakrabarti looped in the product manager overseeing the XCheck program to offer a response.

The fairness concerns were real and XCheck had been mismanaged, the product manager wrote, but “we have to balance that with business risk.“ Since the company was already trying to address the program’s failings, the best approach was “internal transparency,” he said.

On May 5, Facebook’s Oversight Board upheld the suspension of Mr. Trump, whom it accused of creating a risk of violence in connection with the Jan. 6 riot at the Capitol in Washington. It also criticized the company’s enforcement practices, recommending that Facebook more clearly articulate its rules for prominent individuals and develop penalties for violators.


In May, Facebook’s Oversight Board upheld the suspension of former President Donald Trump.
PHOTO: ANDREW HARRER/BLOOMBERG NEWS


As one of 19 recommendations, the board asked Facebook to “report on the relative error rates and thematic consistency of determinations made through the cross check process compared with ordinary enforcement procedures.”

A month later, Facebook said it was implementing 15 of the 19 recommendations. The one about disclosing cross check data was one of the four it said it wouldn’t adopt.

“It’s not feasible to track this information,” Facebook wrote in its responses. “We have explained this product in our newsroom,” it added, linking to a 2018 blog post that declared “we remove content from Facebook, no matter who posts it, when it breaks our standards.” Facebook’s 2019 internal review had previously cited that same blog post as misleading.

The XCheck documents show that Facebook misled the Oversight Board, said Kate Klonick, a law professor at St. John’s University. The board was funded with an initial $130 million commitment from Facebook in 2019, and Ms. Klonick was given special access by the company to study the group’s formation and its processes.

“Why would they spend so much time and money setting up the Oversight Board, then lie to it?” she said of Facebook after reviewing XCheck documentation at the Journal’s request. “This is going to completely undercut it.”

In a written statement, a spokesman for the board said it “has expressed on multiple occasions its concern about the lack of transparency in Facebook’s content moderation processes, especially relating to the company’s inconsistent management of high-profile accounts.”

Facebook is trying to eliminate the practice of whitelisting, the documents show and the company spokesman confirmed. The company set a goal of eliminating total immunity for “high severity” violations of FB rules in the first half of 2021. A March update reported that the company was struggling to rein in additions to XCheck.

“VIP lists continue to grow,” a product manager on Facebook’s Mistakes Prevention Team wrote. She announced a plan to “stop the bleeding” by blocking Facebook employees’ ability to enroll new users in XCheck.

One potential solution remains off the table: holding high-profile users to the same standards as everyone else.

“We do not have systems built out to do that extra diligence for all integrity actions that can occur for a VIP,” her memo said. To avoid making mistakes that might anger influential users, she noted, Facebook would instruct reviewers to take a gentle approach.

“We will index to assuming good intent in our review flows and lean into ‘innocent until proven guilty,’ ” she wrote.

The plan, the manager wrote, was “generally” supported by company leadership.


source
« Last Edit: September 14, 2021, 05:33:40 PM by javajolt »