## Full Text Meta’s platforms are built to be places where people can express themselves freely. That can be messy. On platforms where billions of people can have a voice, all the good, bad and ugly is on display. But that’s free expression. In his 2019 speech at Georgetown University, Mark Zuckerberg argued that free expression has been the driving force behind progress in American society and around the world and that inhibiting speech, however well-intentioned the reasons for doing so, often reinforces existing institutions and power structures instead of empowering people. He said: “Some people believe giving more people a voice is driving division rather than bringing us together. More people across the spectrum believe that achieving the political outcomes they think matter is more important than every person having a voice. I think that’s dangerous.” In recent years we’ve developed increasingly complex systems to manage content across our platforms, partly in response to societal and political pressure to moderate content. ==This approach has gone too far. As well-intentioned as many of these efforts have been, they have expanded over time to the point where we are making too many mistakes, frustrating our users and too often getting in the way of the free expression we set out to enable. Too much harmless content gets censored, too many people find themselves wrongly locked up in “Facebook jail,” and we are often too slow to respond when they do. == > [!note] > I don't think there is much to argue about the assertion that Facebook/Meta got a lot wrong. I do find the change from holding up accuracy rates, to now admitting mistakes interesting. We want to fix that and return to that fundamental commitment to free expression. Today, we’re making some changes to stay true to that ideal. ## Ending Third Party Fact Checking Program, Moving to Community Notes When we launched our independent fact checking program in 2016, we were very clear that we didn’t want to be the arbiters of truth. We made what we thought was the best and most reasonable choice at the time, which was to hand that responsibility over to independent fact checking organizations. The intention of the program was to have these independent experts give people more information about the things they see online, particularly viral hoaxes, so they were able to judge for themselves what they saw and read. That’s not the way things played out, especially in the United States. ==Experts, like everyone else, have their own biases and perspectives. This showed up in the choices some made about what to fact check and how. Over time we ended up with too much content being fact checked that people would understand to be legitimate political speech and debate. Our system then attached real consequences in the form of intrusive labels and reduced distribution. A program intended to inform too often became a tool to censor.   == > [!note] > This is not wrong, per se. It is the timing of the realisation or the public statement of this realisation (assuming, there was a consensus around this internally already but not stated publicly) that is problematic. ^6739ea We are now changing this approach. We will end the current third party fact checking program in the United States and instead begin moving to a Community Notes program. We’ve seen this approach work on X – where they empower their community to decide when posts are potentially misleading and need more context, and people across a diverse range of perspectives decide what sort of context is helpful for other users to see. We think this could be a better way of achieving our original intention of providing people with information about what they’re seeing – and one that’s less prone to bias. - Once the program is up and running, Meta won’t write Community Notes or decide which ones show up. They are written and rated by contributing users. - Just like they do on X, Community Notes will require agreement between people with a range of perspectives to help prevent biased ratings. - We intend to be transparent about how different viewpoints inform the Notes displayed in our apps, and are working on the right way to share this information. - People can sign up today ([Facebook](https://www.facebook.com/help/contact/1914298425761977), [Instagram](https://help.instagram.com/contact/1223551615403090), [Threads](https://help.instagram.com/contact/1638078013752611)) for the opportunity to be among the first contributors to this program as it becomes available. We plan to phase in Community Notes in the US first over the next couple of months, and will continue to improve it over the course of the year. As we make the transition, we will get rid of our fact-checking control, stop demoting fact checked content and, instead of overlaying full screen interstitial warnings you have to click through before you can even see the post, we will use a much less obtrusive label indicating that there is additional information for those who want to see it.   ## Allowing More Speech ==Over time, we have developed complex systems to manage content on our platforms, which are increasingly complicated for us to enforce. As a result, we have been over-enforcing our rules, limiting legitimate political debate and censoring too much trivial content and subjecting too many people to frustrating enforcement actions.== > [!note] > Again, not entirely misplaced, and something they have been criticised for. ==For example, in December 2024, we removed millions of pieces of content every day. While these actions account for less than 1% of content produced every day, we think one to two out of every 10 of these actions may have been mistakes (i.e., the content may not have actually violated our policies). This does not account for actions we take to tackle large-scale adversarial spam attacks. We plan to expand our transparency reporting to share numbers on our mistakes on a regular basis so that people can track our progress. As part of that we’ll also include more details on the mistakes we make when enforcing our spam policies.== > [!note] > Interesting how they suddenly realise mistakes when it is about supporting a decision they seem to want to take. ^c992e9 ==We want to undo the mission creep that has made our rules too restrictive and too prone to over-enforcement. We’re 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. == > [!note] > At face value, not much wrong this intent. ==We’re also going to change how we enforce our policies to reduce the kind of mistakes that account for the vast majority of the censorship on our platforms. Up until now, we have been using automated systems to scan for all policy violations, but this has resulted in too many mistakes and too much content being censored that shouldn’t have been.== > [!note] callout inserted mid-para > Again, in general, the move to limit opaque/automated decisions is not entirely bad. But, again, it is the self-serving nature and timing that is problematic. ==So, we’re going to continue to focus these systems on tackling illegal and high-severity violations, like terrorism, child sexual exploitation, drugs, fraud and scams. For less severe policy violations, we’re going to rely on someone reporting an issue before we take any action. == > [!note] callout inserted mid-para > In theory, not wrong to say that the focus will be on narrower categories of problematic content. The problem, though, is that the implementation is very likely to fall short. It looks like large buckets of content are going to move into a category of 'less severe violations'. There are already documented instances of Meta platforms not being responsive to user reports. What happens when, by policy, the stated plan is to de-prioritise action. Now, Meta PR will likely say something like there are no active plans to de-prioritise taking action against other categories of content/reports. But that is likely to happen when they are treated as different severity levels. ==We also demote too much content that our systems predict might violate our standards. We are in the process of getting rid of most of these demotions and requiring greater confidence that the content violates for the rest. And we’re going to tune our systems to require a much higher degree of confidence before a piece of content is taken down. As part of these changes, we will be moving the trust and safety teams that write our content policies and review content out of California to Texas and other US locations.== ^af045f > [!note] > 1. This was largely unseen, in that, we didn't know what was being demoted/we couldn't see. So we'll have to wait and see how this actually changes what people see on their timelines / or what types of 'content' benefits. > 2. I don't see how moving the teams does anything for the stated problem of demoting too much content. > 1. Are they saying the teams in California were inept / biased? Because then that's for leadership, like the author of this post to answer for. ^df227a ==People are often given the chance to appeal our enforcement decisions and ask us to take another look, but the process can be frustratingly slow and doesn’t always get to the right outcome. We’ve added extra staff to this work and in more cases, we are also now requiring multiple reviewers to reach a determination in order to take something down.== > [!note] callout inseterted mid-para > Ok, we'll have to see if there are any meaningful differences. Press Releases can promise a lot of things that are never actually delivered. ==We are working on ways to make recovering accounts more straightforward and testing facial recognition technology, and we’ve started using AI large language models (LLMs) to provide a second opinion on some content before we take enforcement actions.== > [!note] > Yeah, all of this sounds problematic. ## A Personalized Approach to Political Content Since 2021, we’ve made changes to reduce the amount of ==civic content== people see – posts about elections, politics or social issues – based on the feedback our users gave us that they wanted to see less of this content. But this was a pretty blunt approach. We are going to start phasing this back into Facebook, Instagram and Threads with a more personalized approach so that people who want to see more political content in their feeds can. > [!note] > I don't know if they have used this term often before this. Before today, the last reference listed on about.fb.com appears to be from 2017 ([google dork link]([https://www.google.com/search?q="civic+content"+site:about.fb.com)). And there appears to be no mention of it on transparency.fb.com ([google dork link](https://www.google.com/search?q="civic+content"+site:transparency.fb.com)). Again, I could be wrong, so this one is more of an open question. We’re continually testing how we deliver personalized experiences and have recently conducted testing around civic content. As a result, we’re going to start treating civic content from people and Pages you follow on Facebook more like any other content in your feed, and we will start ranking and showing you that content based on explicit signals (for example, liking a piece of content) and implicit signals (like viewing posts) that help us predict what’s meaningful to people. We are also going to recommend more political content based on these personalized signals and are expanding the options people have to control how much of this content they see. These changes are an attempt to return to the commitment to free expression that Mark Zuckerberg set out in his Georgetown speech. That means being vigilant about the impact our policies and systems are having on people’s ability to make their voices heard, and having the humility to change our approach when we know we’re getting things wrong. ## Colophon title:: More Speech and Fewer Mistakes | Meta type:: [[full-text]] url:: https://about.fb.com/news/2025/01/meta-more-speech-fewer-mistakes/ date:: [[2025-01-07]] published:: 2025-01-07T12:05:30+00:00