Meta’s content moderation changes closely align with FIRE recommendations

On Tuesday, Meta* CEO Mark Zuckerberg and Chief Global Affairs Officer Joel Kaplan announced sweeping changes to the content moderation policies at Meta (the owner of Facebook, Instagram, and Threads) with the stated intention of improving free speech and reducing “censorship” on its platforms. The changes simplify policies, replace the top-down fact-checking with a Community Notes-style system, reduce opportunities for false positives in automatic content flagging, and allow for greater user control of content feeds. All these changes mirror recommendations FIRE made in its May 2024 Report on Social Media.

Given Meta’s platforms boast billions of users, the changes, if implemented, have major positive implications for free expression online.

FIRE’s Social Media Report

FIRE Report on Social Media 2024

Reports

With as many as 5.17 billion accounts worldwide, social media is the most powerful tool in history for average citizens to express themselves.

Read More

In our report, we promoted three principles to improve the state of free expression on social media:

1. The law should require transparency whenever the government involves itself in social media moderation decisions.
2. Content moderation policies should be transparent to users, who should be able to appeal moderation decisions that affect them.
3. Content moderation decisions should be unbiased and should consistently apply the criteria that a platform’s terms of service establish.

Principle 1 is the only one where FIRE believes government intervention is appropriate and constitutional (and we created a model bill to that effect). Principles 2 and 3 we hoped would enjoy voluntary adoption by social media platforms that wanted to promote freedom of expression. 

While we don’t know whether these principles influenced Meta’s decision, we’re pleased the promised changes align very well with FIRE’s proposals for how a social media platform committed to free expression could put that commitment into practice.

Meta’s changes to content moderation structures

With a candid admission that it believes 10-20% of its millions of daily content removals are mistakes, Meta announced it is taking several actions to expand freedom of expression on the platform. The first is simplification and scaling back of its rules on the boundaries of discourse. According to Zuckerberg and Kaplan:

[Meta is] getting rid of a number of restrictions on topics like immigration, gender identity and gender that are the subject of frequent political discourse and debate. It’s not right that things can be said on TV or the floor of Congress, but not on our platforms. These policy changes may take a few weeks to be fully implemented. 

While this is promising in and of itself, it will be enhanced by a broad change to the automated systems for content moderation. Meta is restricting its automated flagging to only the most severe policy violations. For lesser policy violations, a user will have to manually report a post for review and possible removal. Additionally, any removal will require the agreement of multiple human reviewers.

This is consistent with our argument that AI-driven and other automated flagging systems will invariably have issues with false-positives, making human review critical. Beyond removals, Meta is increasing the confidence threshold required for deboosting a post suspected of violating policy.

Who fact-checks the fact checkers?

Replacing top-down fact-checking with a bottom-up approach based on X’s Community Notes feature may be just about the biggest change announced by Meta. As FIRE noted in the Social Media Report: 

Mark Zuckerberg famously said he didn’t want Facebook to be the “arbiter of truth.” But, in effect, through choosing a third-party fact checker, Facebook becomes the arbiter of the arbiter of truth. Given that users do not trust social media platforms, this is unlikely to engender trust in the accuracy of fact checks.

Zuckerberg similarly said in the announcement that Meta’s“fact checkers have just been too politically biased, and have destroyed more trust than they’ve created.” 

Our Social Media Report argued that the Community Notes feature is preferable to top-down fact-checking, because a community of diverse perspectives will likely be “less vulnerable to bias and easier for users to trust than top-down solutions that may reflect the biases of a much smaller number of stakeholders.” Additionally, we argued labeling is more supportive of free expression, being a “more speech” alternative to removal and deboosting.

We are eager to see the results of this shift. At a minimum, experimentation and innovation in content moderation practices provides critical experience and data to guide future decisions and help platforms improve reliability, fairness, and responsiveness to users.

User trust and the appearance of bias

An overall theme in Zuckerberg and Kaplan’s remarks is that biased decision-making has eroded user trust in content moderation at Meta, and these policy changes are aimed at regaining users’ trust. As FIRE argued in our Social Media Report:

In the case of moderating political speech, any platform that seeks to promote free expression should develop narrow, well-defined, and consistently enforceable rules to minimize the kind of subjectivity that leads to arbitrary and unfair enforcement practices that reduce users’ confidence both in platforms and in the state of free expression online.

We also argued that perception of bias and flexibility in rules encourages powerful entities like government actors to “work the refs,” including through informal pressure, known as “jawboning.”

What is jawboning? And does it violate the First Amendment?

Issue Pages

Indirect government censorship is still government censorship — and it must be stopped.

Read More

Additionally, when perceived bias drives users to small, ideologically homogeneous alternative platforms, the result can damage broader discourse:

If users believe their “side” is censored unfairly, many will leave that platform for one where they believe they’ll have more of a fair shake. Because the exodus is ideological in nature, it will drive banned users to new platforms where they are exposed to fewer competing ideas, leading to “group polarization,” the well-documented phenomenon that like-minded groups become more extreme over time. Structures on all social media platforms contribute to polarization, but the homogeneity of alternative platforms turbocharges it.

These are real problems, and it is not clear whether Meta’s plans will succeed in addressing them, but it is welcome to see them recognized.

International threats to speech

Our Social Media Report expressed concern that the Digital Services Act — the broad EU regulation mandating censorship on social media far beyond what U.S. constitutional law allows — would become a least common denominator approach for social media companies, even in the United States. Mark Zuckerberg seems to announce his intention to do no such thing, stating he planned to work with President Trump to push back on “governments around the world” that are “pushing [companies] to censor more.”

While we are pleased at the implication that Meta’s platforms will seemingly not change their free expression policies in America at the behest of the EU, the invocation of a social media company working with any government, including the United States government, rings alarm bells for any civil libertarian. We will watch this development closely for that reason. 

FIRE has often said — and it often bears repeating — the greatest threat to freedom of expression will always come from the government, and as Zuckerberg himself notes, the government has in years past pushed Meta to remove content.

When the rubber meets the road

Meta’s commitment to promote freedom of expression on its platforms offers plenty of reasons for cautious optimism. 

But we do want to emphasize caution. There is, with free expression, often a large gap between stated intentions and what happens when theory meets practice. As a civil liberties watchdog, our duty is to measure promise against performance.

Take, for example, our measured praise for Elon Musk’s stated commitment to free expression, followed by our frequent criticism when he failed to live up to that commitment. And that criticism hasn’t kept us from giving credit when due to X, such as when it adopted Community Notes. 

Similarly, FIRE stands ready to help Meta live up to its stated commitments to free expression. You can be sure that we will watch closely and hold them accountable.

* Meta has donated to FIRE.