Will Meta’s changes to content moderation work?

Mark Zuckerberg’s announcement this week of changes to Meta’s content moderation policies appeared to primarily be about building trust. Trust among users. Trust among investors. And trust among the incoming Trump administration. “It’s time to get back to our roots around free expression,” Zuckerberg said in his announcement.

While we applaud anything that is generally trying to embolden free expression, will these moves actually do that? We break it down –

Fact-checking

Meta is abandoning the use of independent fact checkers on its platforms (Facebook, Instagram and Threads) and replacing them with X-style “community notes”, where commenting on the accuracy or veracity of posts is left to users. But fact checks by dedicated fact-checking organisations do not work against free expression. As a rule they do not remove, override or stifle existing content. Instead they challenge it and contextualise it. As tech expert Mike Masnick wrote after the announcement: “Fact-checking is the epitome of “more speech”— exactly what the marketplace of ideas demands. By caving to those who want to silence fact-checkers, Meta is revealing how hollow its free speech rhetoric really is.”

On the flipside, as Masnick also points out, professional fact checkers are not always effective. The “people who wanted to believe false things weren’t being convinced by a fact check (and, indeed, started to falsely claim that fact checkers themselves were ‘biased’),” he writes. The notion of “bias” was referenced by Zuckerberg himself, who accused fact-checkers of this.

While the set-up that existed until now has been imperfect, are proposed community notes any better? This is complicated. and there is little evidence to suggest they work to the extent that Zuckerberg claims. Community notes tend to be effective for issues on which there is consensus, because there must be agreement before a note can be added to a post. This means that misleading posts on politically divisive subjects often go unchecked, while some accurate posts can be flagged as untrue if enough people determine it that way. According to MediaWise, a media literacy programme at the Poynter Institute, only about 4% of drafted community notes about abortion and 6% of those on immigration were made public on X.

There is also a big difference between those who are paid (and qualified) to fact-check versus non-professionals and this can be evident in the very logistics. According to X, “in the first few days of the Israel-Hamas conflict, notes appeared at a median time of just five hours after posts were created.” In the online world, where a post can go viral within minutes, hours is a long time, arguably too long.

Content moderation

In addition to getting rid of dedicated fact-checkers, Meta is dialling back its content moderation teams and reducing reliance on filters. The move away from automated content moderation processes is to be welcomed. Due to the complexity of speech and online content sharing – with languages and communities evolving slang, colloquialisms and specific terminology – and the ambiguity over imagery, automated processes do not retain the contextual details or complexity necessary to make consistent and informed decisions.

Mis- and disinformation are problematic standards for content removal too. For instance, satire is commonly presented as fact when obviously false and this a central tenet of protected speech across the globe. Simply removing all posts that are deemed to contain misinformation is not and has not worked.

What is more, censoring misinformation does not address the root cause; removing fake news only temporarily silences those that spread it. It doesn’t demonstrate why the information they are spreading is inaccurate. It may even end up giving conspiracy theorists more reason to believe in their theories by feeling that they are being denied access to information. It can end up undermining trust.

Content moderation isn’t just about removing perceived or real misinformation. It is also about removing posts that propagate hate and/or incite violence. Like with misinformation these have to date been imperfectly applied – sweeping up legal speech and missing illegal speech. Algorithms are ultimately imperfect. They miss nuance and this has had a negative impact on speech across Meta platforms.

It is right for Meta to review these policies as they have too often, to date, failed the free speech test.

Still, in scaling filters back – rather than addressing how to improve them – it does run the risk of allowing a lot more bad content in. Zuckerberg, by his own admission, says that the newly introduced measures are “a “trade-off”. “It means we’re going to catch less bad stuff, but we’ll also reduce the number of innocent people’s posts and accounts that we accidentally take down.”

The flipside of catching “less bad stuff” can be, ironically, less free speech. Harassment can drive people to silence themselves or leave online spaces entirely. This form of censorship (self-censorship) is insidious and cannot be easily measured. Unchecked it can also lead to some of the gravest attacks onto human rights. In 2022 Amnesty issued a report looking into Meta’s role in the Rohingya genocide. It detailed “how Meta knew or should have known that Facebook’s algorithmic systems were supercharging the spread of harmful anti-Rohingya content in Myanmar, but the company still failed to act”.

Following Zuckerberg’s announcement, Helle Thorning-Schmidt, from Meta’s oversight board, said: “We are seeing many instances where hate speech can lead to real-life harm.” She raised concerns about the potential impact on the LGBTQ+ community as just one community.
Another damning response came from Maria Ressa, Rappler CEO and Nobel Peace Prize winner:
“Journalists have a set of standards and ethics. What Facebook is going to do is get rid of that and then allow lies, anger, fear and hate to infect every single person on the platform.”

Finally, Zuckerberg said the remaining content moderation teams will be moved from California to Texas where, he said, “there is less concern about the bias of our teams”. As pointed out by many, including the Electronic Frontier Foundation, there is no evidence that Texas is less biased than California. Due to the political leadership of Texas and the positioning of this state and the perception that it is more closely allied with the incoming administration, there are real concerns that this is replacing one set of perceived biases with another. Instead, a free-speech first approach would be to address what biases exist and how current teams can overcome them, irrespective of geographical location. Establishing a process based on international human rights and free expression standards would be a step in the right direction.

Hateful conduct policy

In Zuckerberg’s announcement he stated “we’re going to simplify our content policies and get rid of a bunch of restrictions on topics like immigration and gender that are just out of touch with mainstream discourse. What started as a movement to be more inclusive has increasingly been used to shut down opinions and shut out people with different ideas, and it’s gone too far.”

Simplifying the policies can increase their efficacy, with users clearer as to the standards employed on the platforms. However, suggesting that policies must move with “mainstream discourse” is a challenging threshold to maintain and could embed uncertainty into how Meta responds to the ever-changing and complex speech environment. Identifying topics such as immigration and gender threatens to define such thresholds by the contentious topics of the day and not objective standards or principles for free expression.

It could also open the floodgates to a lot of genuine hate speech and incitement, which will be incredibly damaging for many individuals and communities – in general and in terms of free speech.
• Foreign interference

In Zuckerberg’s speech he took issue with foreign interference. Platforms and governments have often collided over their interpretations of what is acceptable content and who has the power to decide. Ideally we’d have standardised community guidelines and rules of moderation in line with international human rights law. In practise this is not the case. Except instead of highlighting countries where the human rights record is woeful and content removal requests have been clearly politically motivated, Zuckerberg cited Latin America and Europe here. Article19 said they were “puzzled by Mark Zuckerberg’s assertion that Europe has enacted an ‘ever-increasing number of laws institutionalizing censorship’” and that it showed “misunderstanding”.

Parking a discussion of EU laws, it was certainly disappointing for the reasons stated above. As reported by the Carnegie Center in 2024: “In illiberal and/or autocratic contexts, from Türkiye to Vietnam, governments have exploited the international debate over platform regulation to coerce tech companies to censor—rather than moderate—content.” That is where we need to be having a conversation.

Countries such as India have demonstrated processes by which political pressure can be exerted over content moderation decisions undertaken by social media platforms. According to the Washington Post, the Indian government has expanded its pressure on X: “Where officials had once asked for a handful of tweets to be removed at each meeting, they now insisted that entire accounts be taken down, and numbers were running in the hundreds. Executives who refused the government’s demands could now be jailed, their companies expelled from the Indian market.” Further in the piece, it states: “Records published by the Indian Parliament show that annual takedown requests for posts and accounts increased from 471 to 6,775 between 2014 and 2022, with those to Twitter soaring from 224 in 2018 to 3,417 in 2022.”

Zuckerberg’s announcement was silent on how Meta would respond to or resist such explicit state censorship in countries with weak and eroding democratic norms and standards.

Final thoughts

For now Meta says it has “no immediate plans” to get rid of its third-party fact checkers in the UK or the EU, nor could it necessarily do so because of the legal landscape. Some countries also have outright bans on Meta’s platforms, like China. So this is a story that will play out primarily in the USA.

Still, it is part of a broader pattern of Silicon Valley executives misusing the label “free speech” and the timing of it suggests the motivation is for political gain. Even incoming president Donald Trump acknowledged that this week. The shift towards kowtowing to one party and one person, which we have seen occur on other platforms, is incredibly worrying. As Emily Maitlis said on the News Agents this week when evaluating the announcement: “There is a king on the top here and there are courtiers and they recognise that their position is in terms of how they respond to the king now”.

Whether the platforms are used for sharing pictures of your family or galvanising support for a campaign, we know the powerful and central role social media plays in our lives. Furthermore, according to a 2022 OECD report, around four out of 10 respondents said they did not trust the news media, and more and more people were turning to social media for their news, especially young people. As a result it’s essential that social media lands in a helpful place. Content moderation policies at scale are incredibly difficult and cumbersome. They are impossible to do perfectly and easy to do badly. Still, we have little faith that these changes will be helpful and concerns that they could be hurtful.

We will continue to monitor the situation closely. In the meantime, please do support organisations like Index who are genuinely dedicated to the fight against censorship and the fight for free expression.

The post Will Meta’s changes to content moderation work? appeared first on Index on Censorship.